List of Archived Posts

2005 Newsgroup Postings (08/14 - 08/31)

The Chinese MD5 attack

The Chinese MD5 attack

X509 digital certificate for offline solution

The Chinese MD5 attack

Robert Creasy, RIP

Code density and performance?

X509 digital certificate for offline solution

X509 digital certificate for offline solution

Non Power of 2 Cache Sizes

Need a HOW TO create a client certificate for partner access

Virtual memory and memory protection

ISA-independent programming language

30 Years and still counting

RFC 2616 change proposal to increase speed

dbdebunk 'Quote of Week' comment

open-ssl-stream hangs on post

ISA-independent programming language

Smart Cards?

Data communications over telegraph circuits

Data communications over telegraph circuits

Why? (Was: US Military Dead during Iraq War

help understand disk managment

help understand disk managment

Collins C-8401 computer?

is a computer like an airport?

auto reIPL

How good is TEA, REALLY?

Data communications over telegraph circuits

Penn Central RR computer system failure?

Penn Central RR computer system failure?

auto reIPL

Is symmetric key distribution equivalent to symmetric key generation?

What ever happened to Tandem and NonStop OS ?

Is symmetric key distribution equivalent to symmetric key generation?

Not enough parallelism in programming

Implementing schedulers in processor????

Penn Central RR computer system failure?

What ever happened to Tandem and NonStop OS ?

SHARE reflections

JES unification project

Certificate Authority of a secured P2P network

Certificate Authority of a secured P2P network

Catch22. If you cannot legally be forced to sign a document etc - Tax Declaration etc etc etc

What ever happened to Tandem and NonStop OS ?

Intel engineer discusses their dual-core design

Article: The True Value of Mainframe Security

Article: The True Value of Mainframe Security

Article: The True Value of Mainframe Security

The Chinese MD5 attack

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: The Chinese MD5 attack
Newsgroups: sci.crypt,alt.law-enforcement,misc.legal,alt.usenet.kooks
Date: Sun, 14 Aug 2005 10:09:53 -0600

"Luc The Perverse" writes:

Personally I think the only way to use English language efficiently
in a password is to have words randomly chosen from a dictionary,
with randomly applied word perversion like changing e's to 3's t's
to 7's etc and inserting punctuation in between

If you have any kind of uniform word perversion, and no one else
knows your system, then I suppose a brute force machine would have
no better luck with your password than just random characters.
(Assuming you don't go around advertising I use "system X" to
generate my passwords.)  As long as system X remains something you
don't share, as long as none of your secure passwords are
compromised no one should catch on to the pattern.

a 4/1 password reference from over 20 years ago:
http://www.garlic.com/~lynn/2001d.html#52 OT Re: A beautiful morning in AFM.
a little explanation
http://www.garlic.com/~lynn/2001d.html#51 OT Re: A beautiful morning in AFM.

passwords (and/or passphrases) tend to be some form
of shared-secrets
http://www.garlic.com/~lynn/subintegrity.html#secrets

and because the secret is also exposed at the server-end ... most
security guidelines require that there is a unique shared-secret for
every unique security domain.  (you don't want some temporary
part-time help at the local garage ISP accessing your online bank
account).  this can lead to person having scores of hard to remember
shared-secrets that potentially have to be changed monthly

so there have been various other proposals to address human factors
issues, like using iterative hashing with (single/common) passphrase
and server specific value ... as in RFC 2289, "A One-Time Password
System":
http://www.garlic.com/~lynn/rfcidx7.htm#2289

the above mentions that it is resistant to passive attacker (and is
somewhat related to the subject line since MD5 is one of the possible
hashes).

note, however, a major motivation is to address the proliferation of
unique passwords ... where unique passwords are supposedly
countermeasure against security cross-domain contamination/exploits
(aka the underlying infrastructure requiring unique authentication is
trying to address an attack by somebody in one security domain against
a different security domain).

so a perfectly reasonable attack (given that a major underlying
characteristic/assumption is the whole issue of security cross-domain
contamination/exploit) is where somebody in one security domain
passively listens for salts transmitted by other security
domains. then in a perfectly valid connection to their server, they
accidently spoof a salt from some other domain with a count of one
... followed by some transmission error indication and then restart
the operation ... but with their own salt.

supposedly the whole point of needing uniquely different
authentication for different security domains ... is the possibility
of cross-domain contamination/exploits ... then if there is a claim
about trying to solve a fundamental problem related to multiple-domain
operation (the difficulty of a person remembering scores of different,
unique hard to remember passwords) ... it shouldn't create new kinds
of cross-domain attack vulnerabilities (regardless of whether they are
active or passive attacks).

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

The Chinese MD5 attack

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: The Chinese MD5 attack
Newsgroups: sci.crypt,alt.law-enforcement,misc.legal,alt.usenet.kooks
Date: Sun, 14 Aug 2005 12:47:36 -0600

"Luc The Perverse" writes:

Well if you can't remember a different password for every account,
then just have a small application running on your palm/that berry
device/PSP and have it autogenerate entropy based passwords and
store them in an encrypted file.

I admit I use the same password for all of my accounts which I don't
really care about.  (I mean if someone hacks my "myspace" account
it's not that big of a loss.)

I used completely random characters of at least 12 (usually more)
digits for any kind of account which has information like my
address, phone number back accounts, etc.

I feel pretty good about this

there have been some number of pilots where debit-cards are using
fingerprints in lieu of PIN ... and numerous people have pointed out
how easy it is to spoof fingerprint systems.

the basic premise is that PIN (or biometrics) is countermeasure to
lost/stolen card (aka two-factor authentication where the different
factors have different vulnerability characteristics).

the issue is that some surveys claim that 30precent of the population
write their pins on their debit cards.

then it isn't whether or not it is possible to spoof a fingerprint
system ... the issue is whether it is easier for a crook to enter a
pin that is written on a lost/stolen card ... or it is easier for a
crook to lift (the correct) fingerprint from the lost/stolen card and
enter it.

in any case, most security paradigms are designed for large segment of
the population ... and you aren't going to find the segment of the
population that write 4-digit PINs on their debit cards ...
choosing/remembering 12 random characters passwords.

slight drift on exploits/vulnerabilities:
http://www.garlic.com/~lynn/aadsm20.htm#1 Keeping an eye on ATM fraud
http://www.garlic.com/~lynn/aadsm20.htm#23 Online ID Thieves Exploit Lax ATM Security

with respect to referenced phishing attacks, in the past, a
(remembered) PIN represents two-factor authentication as a
countermeasure to a lost/stolen card ... since they have different
vulnerabilities (when the PIN isn't written on the card). an emerging
problem with phishing attacks is that the PIN and account number
(enabling the manufacture of a counterfeit card) can share common
vulnerability. In that respect, fingerprint is still a lot more
difficult to phish over the internet (than it is to phish either pins,
passwords or account numbers).

however, actually using fingerprint authentication over the internet
can represent more risk than using password authentication over the
internet. an evesdropping compromise of password can be remediated by
replacing the compromised password; it is still somewhat more
difficult to replace a finger.

frequently, biometric authentication will be implemented as
shared-secrets paradigm
http://www.garlic.com/~lynn/subtopic.html#secrets

... and it may get difficult to come up with a unique biometric
shared-secret for every unique security domain (as well as replacing a
body part when its biometric value is compromised).

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

X509 digital certificate for offline solution

From: "" <lynn@garlic.com>
Newsgroups: microsoft.public.dotnet.security
Subject: Re: X509 digital certificate for offline solution
Date: Sun, 14 Aug 2005 14:09:03 -0700

Valery Pryamikov wrote:

It's a bit embarrassing for me to admit that until now I didn't even
check the original question ;-). But I don't think it was question
about business process applicability, but rather a sign of complete
misconception. My understanding of original question is that op was
asking about a way of protecting piece information that is used by
some service (daemon) from everyone else using this computer,
including administrator/root (because if it was only about
protecting against unprivileged users of this computers --
simple access control would be more than enough).

Of course PKI is completely irrelevant here!... but any other
encryption related technology is irrelevant here as well... Since
service/daemon requires protected information in clear text, which
means that decryption key must be accessible to that service on that
computer, but that automatically makes this secret key to be
accessible to administrator/root of this computer as well. The op's
problem as it is, is more close to DRM than to anything else
(i.e. store secret key, and cipher text in one place and hope that
nobody will be able to put them together).

remember in my context, i described asymmetric cryptography as
technology and public keys, digital signatures and PKIs as all
business processes
http://www.garlic.com/~lynn/2005n.html#33 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005n.html#43 X509 digital certificate for offline solution

... so a case study from another PKI scenario where the relying-party
is offline and/or doesn't have realtime direct contact with the
certification authority (which somewhat turns out to actually be the
original design point for PKIs ... the offline situation where the
relying-party doesn't have realtime, online and/or local resources for
resolving information regarding first time communication with a
stranger).

crica 1999, one of the major PKI certification authorities approached a
large financial operation and convinced them that they needed to deploy
a PKI infrastructure enabling their customers to do online, internet
account transactions. This was a financial operation that had
significantly in excess of 10 million accounts.

the scenario went:

1) the financial institution would register a customer public key for
every account

2) the financial institution would transmit the resulting updated
account database to the certification authority

3) the certification authority would munge and re-arrange the bits in
each account record ... producing one digital certificate for each
account record.

4) after a couple hrs, the certification authority would return all the
recently produced digital certificates to the financial operation ...
which would then store them in the appropriate account record and
convey a copy to the appropriate customer

5) the customers would then generate digitally signed account
transactions, package the account transaction, the digital signature
and their copy of the digital certificate and transmit it over the
internet to the financial operation.

6) the financial operation would pull the account number from the
transaction, retrieve the corresponding account record, verify the
digital signature using the public key in the account data base ... and
NEVER have to make any reference to the digital certificate at all

the financial operation had spent nearly $50million on integrating a
PKI infastructure when it dawned on somebody to do the complete
financials.

they had already agreed that the certification authority would get
$100/annum/account for the production of (redundant and superfluous)
digital certificates that need NEVER actually be used.

doing the complete financials resulted in somebody realizing that the
financial operation would be paying the certification authority
$100m/annum per million accounts (or $1b/annum per 10 million
accounts) for redundant and superfluous digital certificates
that need NEVER actually be used ... aka certificate-less
operation (other than the internet payload for continously
transmitting the digital certificates hither and yawn)
http://www.garlic.com/~lynn/subpubkey.html#certless

the financial operation eventually canceled the project and took the
$50m hit.

this was actually a relying-party-only certificate scenario
http://www.garlic.com/~lynn/subpubkey.html#rpo

where the operational account contains all the information about the
entity and the entity's public key (as well as copy of the entity's
public key and a stale, static copy of a subset of the entity's
operational information in the form of a digital certificate).

this is offline, from the standpoint of the relying-party not needing
to contact the certification authority when processing a digitally
signed transaction ... in part, because the relying party actually has
all the real-time operational information as part of executing the
transaction (and NEVER actually needs to reference the redundant and
superfluous, stale, static digital certificate).

however, the certification auhtority was originally expecting to be
paid $100m/million-accounts (well in excess of billion dollars) per
annum for the redundant and superfluous, stale, static
(and need NEVER be referenced) digital certificates.

now, a number of operations have used tamper-resistant hardware tokes
(like USB dongles) as repository for protecting the confidentialty of
private keys. This becomes truely a something you have
operation ...  since the hardware tokens perform operations using the
embedded private key ... but the private key never exists outside of
the confines of the token.

human operators and other agents can still compromise any system useage
involving the private keys ... which is an overall system security
integrity issue. however, the private keys are never divulged ...
eliminating the system security confidentiality issue with regard to
the private keys ... and crooks can't obtain the private keys and setup
a counterfeit operation impersonating the original system (possibly
unknown to the original operation).

this is my frequent refrain that most operations treat public key
operation as a something you have authentication ... aka the party
has accesws and use of the corresponding private key. When purely
software implementation is used ... there typically are attempts to
closely emulate real hardware token operation ... however software
emulation of hardware tokens have several more threats, exploits and
vulnerabilities compared to real hardware tokens.

one way of looking at this issue is where does the security perimeter
lay. the security perimeter for a hardware token ... tends to be the
physical space of the token. the security perimeter for a software
emulated hardware token may be all the components of the computer where
that software is running.

for financial environments ... like PIN-debit ... there are external
tamper resistent hardware boxes (frequently referred to as HSMs) that
do all the PINs processing for the financial institution. PIN are
entered in the clear at POS terminals and ATMs ... but then
immediately encoded. From then on, they never appear in the clear. the
backend gets the transaction and sends it to the box .. and gets back
some answers ... but standard operation never sees the PIN in the
clear.

the (security) integrity of the backend systems might be compromised by
insiders ... but you don't find insiders harvesting PINs from the
backend systems (i.e. security confidentiality) and using them in
impersonation attacks with counterfeit transactions.

part of this is that security integrity compromises tend to be a lot
more difficult than security confidentiality compromises (copying the
data). security integrity compromises also tend to leave a lot more
traces back to the people responsible ... compared to security
confidentiality compromises.

one of the claims that we've frequently made with respect to aads chip
strawman
http://www.garlic.com/~lynn/x959.html#aads

for public key operation, was that being free from having to worry
about the ins & outs of PKIs and digital certificates .... we were
therefor able to concentrate on the fundamental threats and
vulnerabilities of the actual operation of public key processes.

For instance, a very fundamental threat and vulnerability is the
integrity and confidentiality of the private key.  If i was looking at
dieing to spend $100/annum on stuff associated with public key
operation ... I might consider it much better spent on hardware tokens
than on digital certificates.

in fact in the security proportional to risk scenarios ... slightly
related
http://www.garlic.com/~lynn/2001h.html#61

i.e. we've take it as a given that the integrity of the
originating/remote environment is taken into account when evaluating a
transaction for authorization. this includes the risk level associated
with whether or not a real hardware token is being used and if a real
hardware token is being used ... the evaluated, integrity of that
token (which might change over time as technology changes). For a
large percentage of the business processes in the world, we assert
that the integrity level of the remote end is of more importance the a
lot of personal information about the remote entity (which the vast
majority of operations already have on file ... so it is of little
interest to duplicate such information in digital certificates).

so another simple test ....

i would assert that the integrity level of the originating environment
(software token or hardware token and the assurance level of such
token) is one of the primary pieces of information that would be of
interest to a relying-party ... right up there with what is the public
key. so a real buisness oriented digital certificate would not only
give the public key ... but also provide the integrity level of the
environment protecting the private key.

when i examined x.509 fields several years ago ... i couldn't find one
that provided the integrity level of the private key protection
although some have simple flag that can be used to indicate whether it
is software private key or hardware token private key. how many
certification authorities have you heard of that have a process of
checking whether they are certifying a software private key or a
hardware token private key?

the literature has lots of stuff about the integrity level of
public/private keys based on the number of bits in the key ... but I
haven't seen anything on the integrity level of private key protection
... and/or writeups on applications that even make decisions based on
whether they find they are dealing with a software private key or a
hardware private key.

another indication was a couple years ago, i was giving a talk on the
importance of the privatey key protection integrity level ... and
somebody in the audience (from some gov. agency) said that if i would
provide the full definition they would see that it was added to
x.509v3 standard.

The Chinese MD5 attack

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: The Chinese MD5 attack
Newsgroups: sci.crypt,alt.law-enforcement,misc.legal,alt.usenet.kooks
Date: Sun, 14 Aug 2005 16:03:33 -0600

"Luc The Perverse" writes:

It would be better if the authentication was outside of the computer.
I mean if the biometric device in some way had firmware and ran a
program to encrypt and pass on the authentication.   Of course this
would be cracked, so we would need a way to update and patch this as
time went on.

here is a recent two-part posting ... the first part about
pki & digital certificates, the second on integrity levels
of authentication envrionments (specifically integrity levels
related to private keys):
http://www.garlic.com/~lynn/2005o.html#2 X509 digital certificate for offline solutions

one scenario for the aads chip strawman
http://www.garlic.com/~lynn/x959.html#aads

is that PINs become secrets rather than shared-secrets in conjucntion
with digital signature authentication.

a hardware token does on-chip key-gen, exports the public key and the
private key is never divulged. the public key is registered at
manufacturing time along with the integrity level of the chip.
relying-parties that register the public key can obtain the
corresponding chip integrity level from the manufactur (along with any
changes if they are really interested) as well as operational
characteristics of the chip ... i.e. is pin/password &/or biometrics
required for chip operation.

nominally, a relying party verifying a digital signature assumes
single factor something you have authentication ... aka the
responsible entity has access to and use of the corresponding private
key.

however, given the appropriate certification of the public key as to
the characteristics of the operating environment of the corresponding
private key ... the relying party may also be confident as to the
integrity level of the chip protecting the private key (some
confidence that it actually originated from that specific physical
device) and whether or not the chip operation required a PIN or
biometric to operation ... aka what is the confidence level of the
something you have authentication as well as whether or not there
was something you know and/or something you are authentication (in
addtion) ... aka countermeasure for lost/stolen token exploit.

this is somewhat touched on in other threads
http://www.garlic.com/~lynn/aadsm20.htm#21 Qualified Certificate Request
http://www.garlic.com/~lynn/2005m.html#37 public key authentication
http://www.garlic.com/~lynn/2005m.html#53 Barcode Email
http://www.garlic.com/~lynn/2005m.html#54 Barcode Email

a couple years ago, I talked about AADS chip strawman at an assurance
panel in the TPM track at IDF .... and commented that over the
previous two years, TPM started looking more and more like AADS chip
strawman.

note that EU FINREAD standard, a consumer market (home pc) smartcard
reader oriented towards financial transactions has external display and
pin-pad (or possibly biometric sensor)
http://www.garlic.com/~lynn/subintegrity.html#finread

minimizing the possibility that the pin/biometric can be skimmed by
some trojan on the PC ... and the value of any transaction being
digitally signed ... is reliably displayed. It can be used for session
authentication (possibly challenge/response scenario), but it is also
oriented towards transaction authentication (especially financial
transaction authentication) ... where the FINREAD terminal has a
miniture display to (reliably) show what is being digitally signed.

which then starts to stray into the digital signature dual-use hazard.

lots of challenge/response oriented session authentication would have
digital signatures applied to what is assumed to be random data
... and the responsible entity never actually examines what is being
signed.

digital signatures have also been defined for use authenticating
transactions ... where there is some implication that what is being
digitally signed ... is approved, authorized, and/or agreed to.

the digital signature dual-use attack involves the threat of a
comprimized system sending valid transaction data disguised as random
challenge/response data.

some past posts on digital signature dual-use vulnerability:
http://www.garlic.com/~lynn/aadsm17.htm#55 Using crypto against Phishing, Spoofing and Spamming
http://www.garlic.com/~lynn/aadsm17.htm#57 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm17.htm#59 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#0 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#1 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#2 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#3 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#4 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#6 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#12 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#13 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#17 should you trust CAs? (Re: dual-use digital signature vulnerability)
http://www.garlic.com/~lynn/aadsm18.htm#32 EMV cards as identity cards
http://www.garlic.com/~lynn/aadsm18.htm#56 two-factor authentication problems
http://www.garlic.com/~lynn/aadsm19.htm#2 Do You Need a Digital ID?
http://www.garlic.com/~lynn/aadsm19.htm#24 Citibank discloses private information to improve security
http://www.garlic.com/~lynn/aadsm19.htm#41 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm19.htm#42 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm19.htm#43 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm20.htm#0 the limits of crypto and authentication
http://www.garlic.com/~lynn/aadsm20.htm#28 solving the wrong problem
http://www.garlic.com/~lynn/2005.html#14 Using smart cards for signing and authorization in applets
http://www.garlic.com/~lynn/2005b.html#56 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005e.html#31 Public/Private key pair protection on Windows
http://www.garlic.com/~lynn/2005g.html#46 Maximum RAM and ROM for smartcards
http://www.garlic.com/~lynn/2005l.html#20 Newsgroups (Was Another OS/390 to z/OS 1.4 migration
http://www.garlic.com/~lynn/2005m.html#1 Creating certs for others (without their private keys)
http://www.garlic.com/~lynn/2005m.html#11 Question about authentication protocols

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Robert Creasy, RIP

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Robert Creasy, RIP
Date: Mon, 15 Aug 2005 10:35:05 -0600
Newsgroups: bit.listserv.vmesa-l

I recently got a note that Bob Creasy, one of the originators of virtual
machines, died in an accident last week.

You can find references to his contributions, liberally sprinkled in
Melinda's VM history document
http://www.princeton.edu/~melinda/

... a few from above

In the beginning was CTSS, the ``Compatible Time-Sharing
System''. CTSS was written by a small group of programmers at the
Massachusetts Institute of Technology (MIT) in Cambridge,
Massachusetts, under the leadership of Professor Fernando Corbato. One
of the CTSS programmers was Robert Creasy, who was later to become the
leader of the CP-40 project.

....

Rasmussen's response to all this was to decide that the Cambridge
Scientific Center would write a time-sharing system for System/360.
Meanwhile, inside Project MAC, Bob Creasy was upset by the inability
of his colleagues to come to terms with IBM. He was impressed by the
promise of machine upward compatibility offered by S/360, and he
wanted Multics to be a mainstream system. When he heard that Rasmussen
intended to build a time-sharing system based on S/360 and needed
someone to lead the project, Creasy also left MIT to move to the
Scientific Center. Inside IBM, losing the Project MAC bid was
immediately recognized as a serious problem. A corporate task force
was formed to get the company into a position to be able to win bids
for time-sharing systems. The task force was composed of the most
knowledgeable time-sharing people from around the company. CSC was
represented by Rasmussen, Harmon, Creasy, and Comeau.

....

Creasy had, of course, spotted the most important aspect of the
System/360 announcement, that programs written for one model of S/360
would run on any other model as long as they contained no
timing-dependent code. From the System/360 ``blue letter'' (April 7,
1964): Whatever your customer's data handling requirements are now,
whatever they will be in the foreseeable future, the System/360 can be
custom-fitted to perform his job.  In fact, this amazing new system
makes possible, for the first time in the industry, a truly long-range
growth plan for all customers. For it can accommodate virtually any
combination of processing and computing functions. And can expand in
easy, economical steps as the customer's needs change---with little or
no reprogramming, no retraining of personnel, no disruption of
service.

....

In the Fall of 1964, the folks in Cambridge suddenly found themselves
in the position of having to cast about for something to do next. A
few months earlier, before Project MAC was lost to GE, they had been
expecting to be in the center of IBM's time-sharing activities. Now,
inside IBM, ``time-sharing'' meant TSS, and that was being developed
in New York State. However, Rasmussen was very dubious about the
prospects for TSS and knew that IBM must have a credible time-sharing
system for the S/360. He decided to go ahead with his plan to build a
time-sharing system, with Bob Creasy leading what became known as the
CP-40 Project.

....

What was most significant was that the commitment to virtual memory
was backed with no successful experience. A system of that period that
had implemented virtual memory was the Ferranti Atlas computer, and
that was known not to be working well.  What was frightening is that
nobody who was setting this virtual memory direction at IBM knew why
Atlas didn't work. Creasy and Comeau spent the last week of 1964
joyfully brainstorming the design of CP-40, a new kind of operating
system, a system that would provide not only virtual memory, but also
virtual machines.

....

Code density and performance?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Code density and performance?
Newsgroups: comp.arch
Date: Mon, 15 Aug 2005 14:00:49 -0600

Seongbae Park writes:

For example, Sun's Performance Analyzer can produce a mapfile from
performance measurement to optimize for code locality (and it's
certainly not the only one that can do that sort of thing).  This
mapfile can be used in your build environment - it requires only a
minimal impact in the build environment since the mapfile doesn't
need to be generated everytime you build.

note that vs/repack did that ... it was done at the science center
http://www.garlic.com/~lynn/subtopic.html#545tech

in the early 70s ... and was used as part of the analysis of
rewritting apl storage maangement (garbage collection) in the port
from apl\360, small (16kbytes to 32kbytes) real storage swapping to
cms\apl large virtual memory operation.

it was also used for analysis of some of the "large" 360 application
in moving them from the os/360 real storage environment to os/vs2
virtual memory environment.

it was released as a product in the March of 1976.

i was a little annoyed ... i had written some of the data collection
software for vs/repack ... but the real analysis work was done by
hatfield. it was priced software and the science center was on the
list of organizations ... that if employees produced priced software,
they got the equivalent of one month license for all copies sold the
first year.

in april '76, the science center was removed from the list of
organizations, where employees got the first month's license.

i got to release the resource manager in may of '76 ... which was the
first priced operating system code ... aka the unbundling announcement
of 6/23/69 started pricing for application software, but kernel
software was still free. the resource manager got chosen to be the
guinea pig for priced kernel software (i got to spend a couple months
with business people working out kernel pricing policty) ... lots of
past posts on unbundling and resource manager being priced
http://www.garlic.com/~lynn/subtopic.html#unbundle

i even offered to give up my salary ... since the first year sales of
the resource manager was so good that the first months license was
well over $1m.

in any case article by Hatfield from the early 70s:

D. Hatfield & J. Gerald, Program Restructuring for Virtual Memory, IBM
Systems Journal, v10n3, 1971

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

X509 digital certificate for offline solution

From: lynn@garlic.com
Newsgroups: microsoft.public.dotnet.security
Subject: Re: X509 digital certificate for offline solution
Date: Tue, 16 Aug 2005 07:33:09 -0700

Valery Pryamikov wrote:

wow!

This is really interesting case study that clearly demonstrated
problem of misplaced trust and dirty play of that "one of the major
certification authorities"...

For all means, financial institution could have set up their own CA
for that project... and of course blind signatures could pass better
for relying-party-only scenarios... but I don't know about their
options to license blind signatures at that time ( great that the
patent expired now

ref:
http://www.garlic.com/~lynn/2005n.html#33 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005n.html#43 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#2 X509 digital certificate for offline solution

but an extremely fundamental characteristic is that setting up a
certification authority serves no useful purpose.

it wasn't a case of misplaced trust ... there was absolutely no
requirement for trust propagation (represented by the existance
of a digital certificate, certification authority, and PKI).

an extremely fundamental principle of PKIs and certification authority
is a trust concept, involving trust propagation, where one party is
certifying information for the benefit of other parties.

this was purely a digital signature authentication situation

it did not involve any requirement for the financial infrastructure to
certify any information for the benefit for any other institution.

there is no other institution involved.

a certification authority is able to proove something ... typically by
verifying the information with the authoritative agency for the
information being certified. PKI and digital certificates encapsulate
the information that is being certified ... creating a trust
propagation information where the information can be trusted by other
parties.

in this situation the financial infrastructure is the authoritative
agency for the information ... they don't need anybody else ... or
even themselves, creating a stale, static certification of some
information that they already know.

i've repeatedly asserted that digital certificates are analogous to
the "letters of credit" from the sailing ship days (a bank certifying
some information for other parties).  In this particular case, we
would have a bank president writting a "letter of credit" addressed to
themself ... say, giving the person's current account balance (at the
time the letter was written). The next time the person comes into the
bank to talk to the bank president, they have to present the "letter
of credit" that the bank president had written to themself, listing
the person's stale, static bank balance at the time the letter was
written. The bank president then approves withdrawals against the
account (w/o checking the current account balance) based on each
individual withdrawal being less than the account balance listed in
the letter of credit that the bank president wrote to themself
possibly a year earlier.

....

all they needed was basic digital signature authentication. there was
no requirement for the financial infrastructure to certify any
information for the benefit of any other institution ... and therefor
no requirement for a certificate authority generating digital
certificates which represent the certification of some information by
one institution for the benefit of some other party.

the purpose of a digital certificate is that one party certifies some
information for the benefit of some other party.

the financial operation is the authoritative agency for the
information that they are interested in ... every transaction has to
reference the account record that is the master copy of all the
information.

a stale, static digital certificate containing a small subset of
information in the account record is redundant and superfluous. in
fact, the digital certificate only contained a database lookup
reference to the actual information ... which is also contained in the
actual account transaction message, making any digital certificate not
only redundant and superfluous but also serves no useful purpose.

the technology is asymmetric key cryptography, what one key (of a
key-pair) encodes, the other key decodes (differentiating from
symmetric key cryptography where the same key both encrypts and
decrypts).

a business process is defined called public key ... where one key is
identified as "public" and made freely available. the other key is
identified as "private", kept confidential and never divulged.

a business process is defined called digital signature. a hash of the
message is calculated and encoded with the private key. the recipient
recalculates the hash, decodes the digital signature with the
corresponding public key and compares the two hashes. if the two
hashes are the same, then the recipient can assume:

1) that the message hasn't changed since the digital signature was
originally calculated

2) something you have authentication, i.e. the origin entity had
access to and use of the corresponding private key.

...

note that blind signatures allow for relying party to know something
is true w/o actually needing to bind it to some entity. the issue for
account-based transactions is that you actually want to authenticate
that the originating entity is an entity authorized to execute
transactions against the account.

blind signatures show up in situations like anonymous, electronic
cash. an anonymous person spends some electronic cash ... the
relying party wants to know that the electronic cash is valid w/o
needing to establish anything about the spending entity.

three factor authentication
http://www.garlic.com/~lynn/subintegrity.html#3factor

• something you have
• something you know
• something you are

for account-based authentication, it doesn't need to establish who you
are ... however, it does need to establish that the entity is an
entity authorized to perform specific account operations.

blind signature schemes eliminate binding to an entity and can be used
where authentication isn't required ... simply validity/truth of the
message.

account-based operations requires (authentication) binding of the
entity to the account ... but can be privacy agnostic ... as in the
case of some of offshore annonymous bank acount ... use of the bank
account is pure authentication w/o any identification. there may or
not be identification to open the account, but only authentication is
needed for using the account (not identification). this is one of the
issues raised in the x9.59 financial transaction standard
http://www.garlic.com/~lynn/x959.html#x959
http://www.garlic.com/~lynn/subpubkey.html#privacy

some blind signature references:
http://citeseer.ist.psu.edu/pointcheval96provably.html
http://en.wikipedia.org/wiki/Blind_signature
http://www.tiscali.co.uk/reference/dictionaries/computers/data/m0051357.html
http://www.rsasecurity.com/rsalabs/node.asp?id=2339
http://www.emergentchaos.com/archives/001511.html

X509 digital certificate for offline solution

From: <lynn@garlic.com>
Newsgroups: microsoft.public.dotnet.security
Subject: Re: X509 digital certificate for offline solution
Date: Tue, 16 Aug 2005 11:59:01 -0700

lynn@garlic.com wrote:

an extremely fundamental principle of PKIs and certification authority
is a trust concept, involving trust propagation, where one party is
certifying information for the benefit of other parties.

a certification authority is able to proove something ... typically by
verifying the information with the authoritative agency for the
information being certified. PKI and digital certificates encapsulate
the information that is being certified ... creating a trust
propagation information where the information can be trusted by other
parties.

in this situation the financial infrastructure is the authoritative
agency for the information ... they don't need anybody else ... or even
themselves, creating a stale, static certification of some information
that they already know.

ref:
http://www.garlic.com/~lynn/2005o.html#6 X509 digital certificate for
offline solution

another fundamental principle of PKIs, certification authorities, and
digital certificates ... is not only is it targeted at trust
propagation ... where the certification authority is certifying some
piece of information for use by other parties ... but the purpose of
the digital certificate is to represent the certified information to
the relying party ... when the relying party

1) has no other means of establishing the validaty of the information
themselves (i.e. the relying party doesn't actually maintain their own
information).

2) is not able to directly contact the authoritative agency
responsible for the information (can't validate the information
themselves and therefor not needing to rely on a certification
authority to validate the information).

3) is not able to directly contact the certification authority
responsible for certifying the information.

aka ... the digital certificate provides certified information trust
propagation when the relying party doesn't have the information
themselves and/or has no way of directly accessing the information
(aka a trust propagation scenario).

repeating the bank president scenario ... and the business process
analogy between digital certificates and letters of credit from the
sailing ship days ... the bank president writes the letter of credit
with their right hand ... and then proceeds to transfer the letter of
credit from the right hand to their left hand .... because obviously
there is no other way for a bank president's left hand to know what
the bank president's right hand is doing.

Non Power of 2 Cache Sizes

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Non Power of 2 Cache Sizes
Newsgroups: comp.arch
Date: Wed, 17 Aug 2005 09:08:02 -0600

"John Mashey" writes:

Sigh.

1) *No one* builds a die with 4MB of cache and only uses 3MB.  Any
designer who proposed that would be fired.  There are different ways to
do this, of which some have been used since the mid-1990s or earlier in
microprocessors, but they involve adding a just a modest amount of
extra silicon to the cache to improve yield.

walk down memory lane ...

big change from 370/168-1 to 370/168-3 was to double the cache size
from 32kbytes to 64kbytes ... however they used the 2kbit to index the
additional entries.

the problem was that worked when the software was running 4k virtual
pages ... but if the software was running with 2k virtual pages, it
would only use 32kbytes.

also, if you had software that switched betweek 2k virtual pages and
4k virtual pages ... the cache was flushed at the switch.

the dos/vs and vs1 operating systems ran 2k virtual pages, mvs ran 4k
virtual pages. vm for "real" address virtual machines ran 4k virtual
pages by default ... but if they were running guest operating systems
that used virtual memory ... the shadow pages were whatever the guest
was using.

there was some large customer running vs1 production work under vm on
168-1 and installed the 168-3 upgrade (32k->64k cache) expecting
performance to get better ... but things got much worse. first, since
vs1 was running 2k virtual pages ... all of the vm shadow tables for
vs1 were 2k virtual pages ... which met the cache only worked with
32kbytes. the other was that vm supervisor reloaded 4k virtual page
mode by default ... so there was lots of cache flush overhead
constantly switching back and forth betweek 2k and 4k virtual page
moades.

the 370 was purely 24bit/16mbyte virtual address spaces. the 168s used
the 8mbyte bit in TLB entry selection. MVS (& svs) had 16mbyte space
mapped out so that there was an 8mbyte MVS kernel image in each of its
virtual address spaces. That left only 8mbytes (in each virtual
address space) for running application (although as system functions
grew ... there were some environments where only 3mbytes in each
virtual address space left for application execution).

the 168 had a 128 entry TLB with seven entry "STO" stack (segment
table origin, basically uniquely identified each virtual address
sapce).

For basic MVS, half the TLB entries went to the kernel image pages and
half the TLB entries went to application pages.

however, a lot of virtual address space evivronments running under VM
tended to start virtual address space use at zero and grow upwards, a
large number of applications rarely ever exceeded a couple mbytes ...
so frequently at least half the TLB entries (for virtual address
>8mbytes) went unused.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Need a HOW TO create a client certificate for partner access

From: <lynn@garlic.com>
Newsgroups: microsoft.public.windows.server.security,microsoft.public.sharepoint.portalserver
Subject: Re: Need a HOW TO create a client certificate for partner access
Date: Wed, 17 Aug 2005 15:24:24 -0700

Serge Ayotte wrote:

Hello to all... If the following is part of a white paper or other
'net resource, please direct me to it... After many trial on Google, I
can't seem to hit the right word combination :)
Never having touch (for not needing it) certificate/pki, I am a bit
lost now...

I need to secure a SharePoint Portal Server site that will be access
from the "outside" of the network (this SPS is running on a member
server of a SBS2K3 domain). I already have SSL and permit only that,
but I was asked to look into a "stronger" method of making sure that
the people accessing it are allowed.
Aside from user/password I immediately thought that a client
certificate would be the way to go, but now I am a bit lost in all the
information I got from MS and Technet about client certificate.

To give a certificate to someone to import on his computer, do I HAVE
to have him connect to a Certificate server site for that? Isn't there
a way for me to be able to generate the certificate and send it to the
"outside partner" I want to give access to the site?

If I am correct, I could then associate the certificate to a login
account, so in a way having a double security level (i.e.
username/password not enough to access, or only the certificate not
enough also).

Thank you in advance for any and all hints, tips, trick and direction
you will provide... Very much appreciated in advance!

from an administrative standpoint ... get a server that supports
RADIUS authentication .... it is probably the most pervasive
authentication methodology on the internet today ... being extensively
deployed by ISP and large number of other organizations. For instance,
if you have ever setup a computer for PPP/dial-in access to an ISP
... typically there has been a screen where you select one of 3-4
different authentication mechanisms ... this is typically then what
your ISP or corporate datacenter has pre-specified for your particular
account in a RADIUS infrastructure.

In addition to authentication, RADIUS also provides additional
optional capability for supporting authorization, permissions, and
accounting on an account by account basis.

RADIUS supports a number of different authentication paradigms ...
having originally started with userid/password ... but there are
versions that have been extended with other types of authentication
methodologies ... where you can actually select the authentication
mechanism on a account by account basis (or userid by userid).

One authentication mechanism is recording public keys in lieu of
passwords and doing digital signature verification
http://www.garlic.com/~lynn/subpubkey.html#radius

this is using the registration of public keys, on file in the radius
infrastructure for performing digital signature verification w/o
requiring PKIs, certification authorities, and/or digital
certificates.
http://www.garlic.com/~lynn/subpubkey.html#certless

the basic technology is asymmmetric key cryptography ... where what
one key (of a key-pair) encodes, the other key (of the key-pair)
decodes.  This is in contrast to symmetric key cryptography where the
same key is used for both encryption and decryption.

a business process is defined called public key, where one of the
asymmetric key pair is identified/labeled "public" and freely
disclosed. The other of the key pair is identified/labeled "private"
is kept confidential and never disclosed.

a business process is defined called digital signature. a hash of a
message or document is calculated and encoded using the private key,
yielding the digital signature. the message is combined with the
digital signature and transmitted. the recipient recalculates the hash
on the message, decodes the digital signature with the corresponding
public key and compares the two hashes. if the two hashes are equal,
then the recipient can assume that

1) the message hasn't be modified in transit
2) something you have authentication, aka the sender has access to
and use of the corresponding private key.

this is slightly modified for pure authentication ... using a
challenge/response protocol. The server sends the client some random
data as a challenge (as countermeasure to replay
attacks). The client calculates the digital signature for the
challenge and returns just the digital signature (since the server has
the challenge). The server calculates the challenge hash, decodes the
client's digital signature that was returned and compares the two
hashes.

there are various kinds attacks that a server and/or imposter may
mount on a client. as countermeasure for some of these attacks
... the client actually adds some of their own random data to the
challenge before calculating the digital signature. the client then
returns both their added data and their digital signature to the
server. the server now has to calculate the hash against a combination
of the original challenge and the added data provided by the client.

At its basic there is no actual need to generate a client digital
certificate and/or require a PKI and/or certification authority. The
basic requirement for a certification authority is to certify the
validaty of some information (represented by the contents of a digital
certificate) for the benefit of other parties which have no means of
otherwise obtaining information about the party they are dealing with.
This is the first time message/communication received from a total
stranger scenario.

Fundamentally all that is needed is for the client to
1) generate a public/private key pair
2) be able to register public key with some server infrastructure
3) be able to generate digital signature with their private key

and for a little drift, one of the possible digital signature attacks
involves dual-use vulnerability involving digital signatures. there
are many instances where digital signatures are used for pure
authentication ... where the digital signature is applied to purely
random data ... that is never actually examined by the signing human.

however, there are also infrastructures where real messages and/or
documents are digitally signed, carrying with it a connotation similar
to that of a human signature, aka that the person has read,
understood, agrees, approves, and/or authorizes what has been
digitally signed. so one possible vulnerability is for an attacker to
transmit to a client a valid contract or financial transaction, under
the ruse of random challenge data. The client then automatically
digitally signs the "random challenge data" w/o ever examining what is
being digitally signed.

misc. past dual-use digital signature vulnerability postings:
http://www.garlic.com/~lynn/aadsm17.htm#25 Single Identity. Was: PKI International Consortium
http://www.garlic.com/~lynn/aadsm17.htm#55 Using crypto against Phishing, Spoofing and Spamming
http://www.garlic.com/~lynn/aadsm17.htm#57 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm17.htm#59 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#0 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#1 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#2 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#3 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#4 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#6 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#12 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#13 dual-use digital signature vulnerability
http://www.garlic.com/~lynn/aadsm18.htm#17 should you trust CAs? (Re: dual-use digital signature vulnerability)
http://www.garlic.com/~lynn/aadsm18.htm#32 EMV cards as identity cards
http://www.garlic.com/~lynn/aadsm18.htm#56 two-factor authentication problems
http://www.garlic.com/~lynn/aadsm19.htm#2 Do You Need a Digital ID?
http://www.garlic.com/~lynn/aadsm19.htm#24 Citibank discloses private information to improve security
http://www.garlic.com/~lynn/aadsm19.htm#41 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm19.htm#42 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm19.htm#43 massive data theft at MasterCard processor
http://www.garlic.com/~lynn/aadsm20.htm#0 the limits of crypto and authentication
http://www.garlic.com/~lynn/aadsm20.htm#28 solving the wrong problem
http://www.garlic.com/~lynn/2004h.html#51 New Method for Authenticated Public Key Exchange without Digital Certificates
http://www.garlic.com/~lynn/2004h.html#58 New Method for Authenticated Public Key Exchange without Digital Certificates
http://www.garlic.com/~lynn/2004i.html#17 New Method for Authenticated Public Key Exchange without Digital Certificates
http://www.garlic.com/~lynn/2004i.html#21 New Method for Authenticated Public Key Exchange without Digital Certificates
http://www.garlic.com/~lynn/2005.html#14 Using smart cards for signing and authorization in applets
http://www.garlic.com/~lynn/2005b.html#56 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005e.html#31 Public/Private key pair protection on Windows
http://www.garlic.com/~lynn/2005g.html#46 Maximum RAM and ROM for smartcards
http://www.garlic.com/~lynn/2005m.html#1 Creating certs for others (without their private keys)
http://www.garlic.com/~lynn/2005m.html#11 Question about authentication protocols
http://www.garlic.com/~lynn/2005o.html#3 The Chinese MD5 attack

Virtual memory and memory protection

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Virtual memory and memory protection
Newsgroups: comp.arch.embedded,comp.arch
Date: Fri, 19 Aug 2005 09:09:17 -0600

"ssubbarayan" writes:

I went through one good tutorial on virtual memory in this website:
http://www.cs.umd.edu/class/spring2003/cmsc311/Notes/Memory/virtual.html

The article says that virtual memory can help in protecting memory of
the applications process.That means each individual process running in
the CPUs memory can be protected.
I am not able to realise how this could help in protection?AFAIK,MMUs
are the ones which help in memory protection and that too MMUS are
implemented in Hardware where you could set some registers to prevent
the processes from crossing boundaries and hence they can protect it.
First of all is my understanding regarding MMUs correct>?
I am not able to understand relation between Virtual memory and memory
protection.Can some one throw some light on this?

I am not from computer science back ground so these things are new to
me.So pardon my ignorance.

class assignment?

standard 360s (in the 60s) were real memory (no virtual memory) and
offered store protect & fetch protect feature.

they also offered supervisor state and problem state ... supervisor
state allowed execution of all instructions (including state change
control and memory ptotect specification). problem state only allowed
execution of a subset of instructions.

store protect feature typically prevented applications from storing
into areas that they weren't allowed to ... like kernel code and/or
other applications.

fetch protect prevented applications from even looking/fetching
regions they weren't suppose to (fetch protect also implied store
protect).

360 model 67 also offerred virtual memory support, it was used to
separate things into totally different domains. not only could
applications be prevented from fetching or storing data into areas
they weren't suppose to (outside of their virtual address space)
... but they possibly weren't even aware that such areas existed. this
was also used to implement virtual machines. recent reference to some
early virtual machine work in the 60s:
http://www.garlic.com/~lynn/2005o.html#4 Robert Creasy, RIP

some number of virtual address space architectures also support r/o
segment protection ... i.e. virtual address space ranges that an
application can't store into. the original 370 virtual memory hardware
was supposed to include r/o segment protect. vm370 was going to use
this to protect common instructions and data areas that could be
shared across multiple different virtual address spaces (only needing
one physical copy of the virtual pages shared across large number of
different application address spaces but still protected). when r/o
segment protect was dropped before product announce ... vm370 had to
fall back to a hack using (360) store protect ... to protect shared
pages. the original relational/sql database implementation was done on
vm370 and made use of various kinds of shared segments
http://www.garlic.com/~lynn/subtopic.html#systemr

a no-execute feature has been added to some hardware recently.  this
is attempting to address the massive number of buffer overruns
vulnerabilities typically associated with c-language implemented
applications. attackers have malicuous executable code introduced via
various kinds of data transfers and then contrive to have execution
transfered to the malicuous code. system facilities use hardware
features to mark pure data areas regions as no-execute. Areas of
(virtual address space) memory that are marked non-executable can be
fetched &/or stored ... but the hardware instruction fetch will
interrupt if instructions are fetched from such a area.

some past postings mentioning no-execute
http://www.garlic.com/~lynn/2004q.html#82 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#0 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#1 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#3 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#5 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#32 8086 memory space [was: The Soul of Barb's New Machine]
http://www.garlic.com/~lynn/2005b.html#25 360POO
http://www.garlic.com/~lynn/2005b.html#39 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005b.html#66 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005c.html#28 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005c.html#44 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005d.html#53 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005d.html#54 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005d.html#55 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005d.html#65 [Lit.] Buffer overruns

some postings mentioning 360 store/fetch protect feature:
http://www.garlic.com/~lynn/2005.html#3 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#5 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005.html#6 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2002f.html#13 Hardware glitches, designed in and otherwise
http://www.garlic.com/~lynn/2003j.html#27 A Dark Day
http://www.garlic.com/~lynn/2005c.html#18 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005c.html#33 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005c.html#47 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005h.html#9 Exceptions at basic block boundaries
http://www.garlic.com/~lynn/2005h.html#17 Exceptions at basic block boundaries
http://www.garlic.com/~lynn/2005j.html#36 A second look at memory access alignment
http://www.garlic.com/~lynn/2005j.html#37 A second look at memory access alignment
http://www.garlic.com/~lynn/2005j.html#39 A second look at memory access alignment
http://www.garlic.com/~lynn/2005l.html#27 How does this make you feel?

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

ISA-independent programming language

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: ISA-independent programming language
Newsgroups: comp.arch
Date: Fri, 19 Aug 2005 14:22:31 -0600

nmm1@cus.cam.ac.uk (Nick Maclaren) writes:

Eh?  I have just checked up to remind myself, and ISO Pascal was only
STARTED about 1977 and wasn't delivered until 1982.  By then, it was
essentially just an IBM PC language and the control of the language
had been, er, taken over by Borland.  Outside that community, it was
more-or-less defunct.

in the 70s, one of the vlsi tools group did a pascal for the
mainframe, which was then used to develop some number of internal vlsi
design tools. it was also released (in late 70s) as a mainframe
product (vs/pascal)and later ported to the rs/6000 (big motivation was
ths internal tools). one of the (external) product guys did sit on the
iso committee.

in the early/mid 90s ... some number of the vlsi tools were spun off
to outside companies (that specialized in vlsi design tools) ..  and a
lot of these tools had to be ported to other (primarily workstation)
platforms.

the mainframe/rs6000 pascal had quite a bit of work done on it
supporting the major (vlsi design tool) production
applications.

trying to get production applications that were possibly 50k-60k lines
of pascal to some of these other platforms ... was a major
undertaking; especially since these pascals appear to have never been
used for anything other than student jobs (and in at least one case,
the organization had outsourced its pascal support, the organization
was maybe 30 minute drive, depending on traffic conditions ... but the
actual support organization was 12 time zones away).

minor past references:
http://www.garlic.com/~lynn/2004d.html#71 What terminology reflects the "first" computer language ?
http://www.garlic.com/~lynn/2004q.html#35 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005e.html#0 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005e.html#1 [Lit.] Buffer overruns

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

30 Years and still counting

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: 30 Years and still counting
Newsgroups: bit.listserv.ibm-main,alt.folklore.computers
Date: Fri, 19 Aug 2005 14:33:58 -0600

steve@ibm-main.lst (Steve Comstock) writes:

I was hoping to host a little party or hospitality event at SHARE,
but I am not able to make it this time (I'm planning to be in
Seattle in March, however). So I thought I'd send a short post to
everyone, thanking them for the ride.

the following is part of a presentation i gave as an undergraduate at
the aug68 share meeting in boston (getting close to 40 years ago).
http://www.garlic.com/~lynn/94.html#18 CP/67 & OS MFT14

I had started with MFT11 sysgen ... taking the stage2 sysgen apart and
breaking it up and re-arranging it so that 1) it could run in the
production jobstream (instead of requiring stand alone time for
starter system) and 2) carefully arraigning files and pds members to
optimize disk arm ... getting about three times speed up for typical
univ. job streams.

in jan. 68 ... three people from the science center had come out and
installed cp/67. in the spring and summer of 68 i rewrote significant
portions of cp/67 kernel reducing cp/67 kernel pathlength time (for
the above referenced jobstream) from 534 cpu seconds. to 113 cpu
seconds.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

RFC 2616 change proposal to increase speed

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: RFC 2616 change proposal to increase speed
Newsgroups: comp.protocols.tcp-ip
Date: Sat, 20 Aug 2005 08:11:05 -0600

dan_delspam writes:

Therefore, my strong suggestion is to increase this value from 2 to
4 like it was already the case vefore in HTTP 1.0.  We could also
put more than 4, sepecially if we define a 20ms delay between each
new connection (to avoid server's load peaks). We could then go up
to easely 10 simultaneous connections, and that would nearly
suppress all delays due to the response time, and therefore improve
a lot the internet's usability.

the issue isn't so much increase speed ... but reduce latency (reponse
time).

For quite a while, I've been using tab browser to handle
differently. i've got a tab bookmark folder with over 100 URLs. I
click on the folder ... and it starts background fetch of all URLs
into different abs.

Get a cup of coffee and then i can "cntrl-pageup" &
"cntrl-pagedown" cycle around the tabs in real time (no latency). I
can click on individuals URL and have them brought into background tab
(latency is being masked while i look at something else).

current issue is somewhere around 250-300 tabs ... the browser starts
to get somewhat sluggish, clicking on a new URL into a background tab
... locks up doing other things in the browser ...  even simple scroll
up/down in the current/active tab.

implications is that there are some browser data structures that are
simple linear lists.

this is analogous, but different to early webserver problems.

tcp was envisioned as being long running sessions. http somewhat
cribbed tcp for simple datagram support. minimum packet exchange for
tcp is seven packets with long dangling finwait when the session was
closed. finwait lists were linear searches. http use of tcp was
resulting in thousands on the finwait list at webservers. the big
webservers were seeing 95% of total cpu use being spent on running
finwait list.

netscape servers were having thruput problems ... so they were being
replicated ... at one point there were something like netscape1 thru
possibly netscape30 ... where users were suggested to type in
different hostname to try and load balance.

as an aside ... i believe it was google (or yahoo?) that did the first
front-end router load-balancing of backend servers ... the google
boundary routers were modified to load balance incoming tcp setup
requests across the backend servers.

at one point, netscape replaced a whole pool of webservers with a
large sequent server. the claim was that possibly sequent had been the
first to rewrite the tcp finwait list handling ... the story was that
they had been selling into the commerical unix market and had
customers with 20,000 concurrent telnet (tcp) sessions into a single
server. With that many long-running telnet sessions ..  there had
started having some significant finwait list activity (http high
finwait list activity isn't from an extremely large number of
concurrent long-running session ... but a high rate of datagram-like,
short-lived requests abusing tcp).

random data point ... vmtp (rfc1045) does reliable transaction
protocol in minimum of 5 packet exchange ... and xtp
http://www.garlic.com/~lynn/subnetwork.html#xtphsp

does reliable transaction protocol in minimum of 3 packet exchange

misc. past posts on tab-browsing
http://www.garlic.com/~lynn/2004e.html#54 Is there a way to configure your web browser to use multiple
http://www.garlic.com/~lynn/2004g.html#6 Adding Certificates
http://www.garlic.com/~lynn/2004h.html#25 Why are programs so large?
http://www.garlic.com/~lynn/2004h.html#51 New Method for Authenticated Public Key Exchange without Digital Certificates
http://www.garlic.com/~lynn/2004i.html#16 New Method for Authenticated Public Key Exchange without Digital Ceritificates
http://www.garlic.com/~lynn/2005c.html#5 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005n.html#8 big endian vs. little endian, why?
http://www.garlic.com/~lynn/2005n.html#15 1.8b2 / 1.7.11 tab performance
http://www.garlic.com/~lynn/2005n.html#41 Moz 1.8 performance dramatically improved

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

dbdebunk 'Quote of Week' comment

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: dbdebunk 'Quote of Week' comment
Newsgroups: comp.databases.theory
Date: Sun, 21 Aug 2005 08:10:11 -0600

"David  Cressey" writes:

Hmmm.  I started out with an IBM 7090 and a PDP-1 (That's a one, not
an eleven), From there moving on to PDP-6 and 10, DECsystem-10 and
20, and VAX.  SQL came much later for me, after Pascal, Datatrieve,
and VAX Rdb/VMS.

recent posting related to 50th anniv share meeting in boston starting
today & having given presentation as undergraduate at the aug68 boston
share meeting
http://www.garlic.com/~lynn/2005o.html#12

part of aug68 presentation
http://www.garlic.com/~lynn/94.html#18

share web page
http://www.share.org/

recent posting mentioning platform original relational/sql
work was done
http://www.garlic.com/~lynn/2005o.html#10

other postings about original relational/sql work
http://www.garlic.com/~lynn/subtopic.html#systemr

and a recent thread that had some pascal discussion
http://www.garlic.com/~lynn/2005o.html#11

slightly older thread discussing vax/vms
http://www.garlic.com/~lynn/2005n.html#10

and even older posting giving number of vax shipments, sliced and
diced by model, year, us, non-us
http://www.garlic.com/~lynn/2002f.html#0

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

open-ssl-stream hangs on post

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: open-ssl-stream hangs on post
Newsgroups: gnu.emacs.gnus
Date: Sun, 21 Aug 2005 08:26:18 -0600

i'm running ngnus 0.3 on fedora FC3 with open-ssl-stream. openssl is
0.9.7a

reading news works fine w/o problem

when i go to post ... the posting process hangs, hanging emacs (same
symptoms with ognus 0.24). initial reaction is to kill emacs and start
over.

however, i also discovered i can kill openssl process and recover.

i find the openssl process and kill it, gnus appears to do a little
processing (nothing displays, emacs remains hung), restarts openssl
process and then stops again.  i find the openssl process and kill it
again. this time gnus finally comes back and says that the posting
failed.

however, the posting actually did work. problem appears to be some
interaction between gnus and openssl regarding whether something
actually happened or not.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

ISA-independent programming language

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: ISA-independent programming language
Newsgroups: comp.arch
Date: Sun, 21 Aug 2005 08:42:00 -0600

re:
http://www.garlic.com/~lynn/2005o.html#11 ISA-independent programming language

the mainframe tcp/ip product was also implemented in vs/pascal.

there were some issues with its thruput ... getting 44kbytes/sec
consuming nearly full 3090 cpu.

i did the product modifications for rfc 1044 support and in some
testing at cray research between a cray and a 4341-clone ... it was
peaking at the channel interface speed using only modest amount cpu
(about 1mbyte/sec using maybe 200-300kips).

past rfc 1044 posts
http://www.garlic.com/~lynn/subnetwork.html#1044

that testing trip sticks in my mind ... we were on nw flight sitting
on the runway at sfo taking off nearly 20 minutes late (for
minneapolis). part way thru the flight, i noticed a number of people
congregated in the back of the plane. i went back to find out what all
the interest was about ... and it turns out they were discussing the
big quake that hit very shortly after we left the ground.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Smart Cards?

From: <lynn@garlic.com>
Newsgroups: microsoft.public.security
Subject: Re: Smart Cards?
Date: Sun, 21 Aug 2005 08:31:14 -0700

Steven L Umbach wrote:

They can be for any domain user. They are a way of using a much more
secure authentication method to logon to a computer that requires a
smart card and pin number. A smart card uses PKI for authentication
and the users smart card contains the users private key. If you
configure a user account to require smart card for logon then there is
no way for someone else to logon as that user unless they have the
users smart card in hand and know the pin number barring a compromise
of the domain, local computer, or PKI such as an unauthorized or
malicious user being or becoming an enrollment agent.

PKI uses what is called a public/private key pair. The public key can
be widely distributed and the private key is considered sensitive and
be protected. A private key can decrypt what a public key encrypts and
can be used as an authentication challenge. Theoretically the
public/private key are unique in that only the one private key can
decrypt what the matching public key can encrypt. The public key is
commonly referred to as the certificate.

the technology is asymmetric key cryptography ... what one key (of a
key-pair) encodes, the other key decodes ... as opposed to symmetric
key cryptography which uses the same key for encryption and decryption.

there is business process called public key ... where one of the keys
(of the key-pair) is labeled as public and freely distributed. the
other key is labeled private, kept confidential and never disclosed.

there is a business process called digital signature ... where the
"signer" computes the hash of some message/document and encodes the
hash with their private key. they then transmit the message/document
and the digital signature. the recipient recomputes the hash on the
message/document, decodes the digital signature with the appropriate
public key (producing the original hash) and compares the two hashes.
if the two hashes are the same, the recipient can assume:

1) the message/document hasn't changed since originally signed

2) something you have authentication ... i.e. the sender has access
to and use of the corresponding private key.

from 3-factor authentication
http://www.garlic.com/~lynn/subintegrity.html#3factor

• something you have (i.e. something kept unique)
• something you know (i.e. pin/password)
• something you are (i.e. biometrics)

recent posting about threats/vulnerability characteristics of
multi-factor authentication
http://www.garlic.com/~lynn/aadsm20.htm#1 Keeping an eye on ATM fraud
http://www.garlic.com/~lynn/aadsm20.htm#23 Online ID Theives Exploit
Lax ATM Security
http://www.garlic.com/~lynn/2005o.html#1 The Chinese MD5 attack

where pin/password (something you know) is frequently used as
countermeasure to lost/stolen object (something you have). one basic
concept of multi-factor authentication ... is the different
factors/mechanisms not having common vulnerabilities.

widely deployed infrastructures for authentication are kerberos
(deployed as basic authentication mechanism in many platforms,
including windows)
http://www.garlic.com/~lynn/subpubkey.html#kerberos
and radius
http://www.garlic.com/~lynn/subpubkey.html#radius

both having started out with userid/password type operation ... where
some sort of shared-secret
http://www.garlic.com/~lynn/subintegrity.html#secret

is registed for authentication. both infrastructures have had upgrades
where public key is registered in lieu of shared-secret and digital
signature authentication occurs in lieu of comparing for password
match. public key having the advantage that it can only be used for
authentication, not impersonation ... where the same shared-secret can
be used for both authentication as well as impersonation.
http://www.garlic.com/~lynn/subpubkey.html#certless

the original internet pk-init draft for kerberos specified simple
public key registration for authentication. it was later extended to
also include the option of using digital certificates.

PKI, certification authorities, and digital certificates are business
processes for addressing the situation involving first time
communication with a stranger. this is the letters of credit model
from the sailing ship days and was targeted at the offline email
scenario from the early 80s where somebody dials their local
(electronic) post office, exchanges email and then hangs up. they are
now potentially faced with first time email from a stranger ... and
they have no local repository or any other means of accessing
information about the stranger.

another issue is that public keys operations are implemented using
both software containers for the something you have private key as
well as hardware tokens. Hardware tokens of various kinds (including
smartcards) are commoningly accepted as having higher integrity level
and trust vis-a-vis software container.

a couple recent postings on subject of integrity levels (and
possibility of the authenticating party needing to determine the
security/integrity level proportional to their risk):
http://www.garlic.com/~lynn/2005o.html#2 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#3 The Chinese MD5 attack

some recent postings about business process characteristics of
certification authorities, digital certificaets, and PKIs.
http://www.garlic.com/~lynn/aadsm20.htm#0 the limits of crypto and
authentication
http://www.garlic.com/~lynn/aadsm20.htm#11 the limits of crypto and authentication
http://www.garlic.com/~lynn/aadsm20.htm#13 ID "theft" -- so what?
http://www.garlic.com/~lynn/aadsm20.htm#15 the limits of crypto and authentication
http://www.garlic.com/~lynn/aadsm20.htm#17 the limits of crypto and authentication
http://www.garlic.com/~lynn/aadsm20.htm#21 Qualified Certificate Request
http://www.garlic.com/~lynn/aadsm20.htm#26 [Clips] Does Phil Zimmermann need a clue on VoIP?
http://www.garlic.com/~lynn/aadsm20.htm#28 solving the wrong problem
http://www.garlic.com/~lynn/aadsm20.htm#30 How much for a DoD X.509 certificate?
http://www.garlic.com/~lynn/aadsm20.htm#31 The summer of PKI love
http://www.garlic.com/~lynn/aadsm20.htm#32 How many wrongs do you need to make a right?
http://www.garlic.com/~lynn/aadsm20.htm#33 How many wrongs do you need to make a right?
http://www.garlic.com/~lynn/2005m.html#37 public key authentication
http://www.garlic.com/~lynn/2005n.html#9 Which certification authority to use
http://www.garlic.com/~lynn/2005n.html#39 Uploading to Asimov
http://www.garlic.com/~lynn/2005n.html#43 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005n.html#49 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#2 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#6 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#7 X509 digital certificate for offline solution
http://www.garlic.com/~lynn/2005o.html#9 Need a HOW TO create a client certificate for partner access

Data communications over telegraph circuits

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Data communications over telegraph circuits
Newsgroups: alt.folklore.computers
Date: Sun, 21 Aug 2005 18:48:26 -0600

Morten Reistad writes:

real world example :

I am currently involved in telephony; and the databases for
subscriptions etc. Telephones have a "high-nine" reliability requirement;
but they are reasonably simple, and the tables to support them are few
and simple too. Interfaces are pretty standard, and good full-systems
failover solutions exist. You can therefore get to 4-5 nines using a
pretty straightforward path.

That is, until you consider the updates and update consistensies to
databases. They have interlocking relationships that make duplication
very difficult.

But, CHANGING a phone subscription is not a five-nine task. OPERATING
IT is. Therefore move changes out, and make the transaction log
handling a separate subsystem; and updating the production facilities
a separate task with a lot lower resiliency than the phones, similar
to the "web stuff" that take subsctiptions.  Then the phones work even
if the database master is blown to smithereens, or the web site is down.

about the samme time we were looking at the previous example
http://www.garlic.com/~lynn/2005n.html#26 Data communications over telegraph circui

we were doing ha/cmp
http://www.garlic.com/~lynn/subtopic.html#hacmp

and we spent some time with a telco organization that had a five-nines
operation. they were currently had a system with triple redundant
hardware connected to a SS7 with replicated T1 links.

the problem was that any (scheduled) software system maint. for the
box consumed close to ten years worth of allowed down time.

replicated high availability system addressed the system down time
issue related to doing software maint. ... and while the individual
hardware boxes might not quite have five-nines ... the SS7 logic
already would redrive request on the 2nd T1 if it didn't get an answer
on the first T1. The SS7 redrive logic would mask any individual
outage ... and while an individual backend box might not be quite
five-nines ... the overall infrastructure (becuase of the SS7 redrive
logic) would be much better than five-nines.

the solution by the replicated hardware group ... was to install a
pair of replicated hardware boxes ... in order to meet the overall
system five-nines objective.

however, once resorting to a pair of backend boxes to handle the
overall system outage issue ... then it was no longer necessary to
have the individual backend boxes be done with replicated hardware
... aka a high availability backend configuration (relying on the SS7
fault masking redrive) made it redundant and superfluous for
having redundant hardware for the individual backend computers.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Data communications over telegraph circuits

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Data communications over telegraph circuits
Newsgroups: alt.folklore.computers
Date: Mon, 22 Aug 2005 08:00:50 -0600

Morten Reistad writes:

Combat systems have the inverse priority BTW. Keep operating
at all costs. An F16 does not have a stable fall-back flying mode;
it must have the computer running. Probably a better example for
your case.

and of course the person responsible for much of the f16 design
http://www.garlic.com/~lynn/subboyd.html#boyd2
http://www.garlic.com/~lynn/subboyd.html#boyd

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Why? (Was: US Military Dead during Iraq War

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Why? (Was: US Military Dead during Iraq War
Newsgroups: alt.folklore.computers
Date: Mon, 22 Aug 2005 09:09:32 -0600

blmblm writes:

Sing it.  The argument against this sort of centralized scheme was
(if I remember right) that it didn't give users enough control over
their own environments, but I rarely found that to be a problem in
the days when my office "computer" was a terminal connected to a
good multiuser shared system.  Install software?  most Unix stuff
installs fine in a user-owned directory, no admin privileges needed.
Maybe it was similar with TOPS-10?

the argument i frequently heard was that the support costs supposedly
disappeared from the budget ... since everybody was mostly doing their
own support ... so it didn't actually show up in a budget that needed
management, reporting, defending, etc.

other business units constantly complaining about how much money was
going to the datacenter ... just eliminate the datacenter and let each
individual business unit eat their own costs.

there was the issue that if you have a computer supporting thousands
of users ... you tended to do a lot more testing and change control
... in case things went wrong. people could willy, nilly do things to
their own computer and (the common opinion at the time), it just
affected the individual.

personal responsibility and empowerment.

misc. time-sharing posts
http://www.garlic.com/~lynn/subtopic.html#timeshare

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

help understand disk managment

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: help understand disk managment
Newsgroups: linux.redhat.misc
Date: Mon, 22 Aug 2005 09:45:07 -0600

"peter" writes:

Thanks, now I know why I was confused. There are two types of file
systems in redhat.

In the simple type, hard drives are partitioned and partitions are
mounted.

In the complicated type (LVM), drives are divided into physical
volume, then physical volumes are combineded into a logical volume
group, which is like a big resizable virtual drive. This logical
volume group can then be "partitioned" into logical volumes which
can then be mounted and resized at will.

Is this more or less correct? What is the correct way to refer to
the simple type file system?

disk partitioning, ala  pc/dos
     ... sort of like minidisks from cp/67 ... circa mid-60s

minor ref
http://www.garlic.com/~lynn/2004e.html#38 [REALLY OT!] Overuse of symbolic constants
http://www.garlic.com/~lynn/2004h.html#40 Which Monitor Would You Pick??????
http://www.garlic.com/~lynn/2004k.html#46 Xah Lee's Unixism

LVM started with aix/v3 for rs/6000
     included support for things other than straight partitioning,
     striping, mirroring, concatenation, etc

as an aside ... journaled file system (JFS) was also done for
for the same rs/6000 aix/v3 release

....

... so one flavor is sort of the pc/dos heritage ... modulo minidisks
and virtual machine stuff from the mid-60.

the other flavor is somewhat more unix heritage ... at least from
aix/v3.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

help understand disk managment

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: help understand disk managment
Newsgroups: linux.redhat.misc
Date: Mon, 22 Aug 2005 12:14:42 -0600

"peter" writes:

Another reason I was confused is caused by the name "logical volume
group".  The name implies it is a group of logical volumes when in
fact it is a group of physical volumes. I suggest we rename it
"physical volume group".

one of the reason for the original LVM label ... were options other
than 1:1 mapping between the logical volume and the physical
partition(s) ... aka mirroing, stripping, concatenation.

the ms/pc dos stuff purely did physical partitioning with a 1:1
mapping between the (virtual) disk (for filesystem) and the physical
partition.

when we did ha/cmp product
http://www.garlic.com/~lynn/subtopic.html#hacmp

... another minor reference
http://www.garlic.com/~lynn/95.html#13

we used lvm mirroring for no-single-point-of-failure ... and JFS for
fast restart after outage.

there were even gimmicks played with 3-way mirroring for backup ...
add a (LVM) 3rd mirror on the fly ... wait until it sync'ed ... and
then took one of the mirrored disks offline and sent offsite for
disaster/recovery.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Collins C-8401 computer?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Collins C-8401 computer?
Newsgroups: alt.folklore.computers
Date: Mon, 22 Aug 2005 13:49:18 -0600

hancock4 writes:

I don't think GE was considered part of the "BUNCH" (Burroughs, Univac,
Control Data, and Honeywell).

GE was one of the seven dwarfs ... and then sold-off computer stuff
to honeywell ... part of reduction from seven to five.

ref. from search engine
http://www.answers.com/topic/honeywell

.. from the above

In the mid-1960s, Honeywell's 200 series gave IBM serious
competition. It outperformed IBM's very successful 1401 computer,
which it emulated, causing IBM to accelerate its introduction of its
System/360. In 1966, Honeywell acquired Computer Control Company's
minicomputer line, and in 1970, it acquired the assets of GE's
computer business. The computer division was renamed Honeywell
Information Systems, Inc.

....

misc. past dwarfs postings
http://www.garlic.com/~lynn/2002o.html#78 Newsgroup cliques?
http://www.garlic.com/~lynn/2003b.html#61 diffence between itanium and alpha
http://www.garlic.com/~lynn/2003.html#36 mainframe
http://www.garlic.com/~lynn/2003.html#71 Card Columns
http://www.garlic.com/~lynn/2003o.html#43 Computer folklore - forecasting Sputnik's orbit with
http://www.garlic.com/~lynn/2004h.html#15 Possibly stupid question for you IBM mainframers...  :-)
http://www.garlic.com/~lynn/2004l.html#19 FW: Looking for Disk Calc program/Exec (long)
http://www.garlic.com/~lynn/2005k.html#0 IBM/Watson autobiography--thoughts on?

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

is a computer like an airport?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: is a computer like an airport?
Newsgroups: comp.protocols.tcp-ip
Date: Mon, 22 Aug 2005 19:04:47 -0600

stuff goes on in different time frames.

airports have flights arriving early, late, delayed ... and gates can
quite frequently get re-assigned ... as airlines & airports
attempt to juggle resources in the face of glitches.

it used to be that planes spent a lot more time circling ... ATC now
tries to hold flights before take-off until there are scheduled
resources available for landing.

delayed in-coming equipment can really start to mess up outgoing
operations ... and the whole infrastructure gets into negative
feedback ... only thing apparently saving it the overnight slowdown.

i've frequently described a flight from san jose to boston with
connection in chicago. the san jose flight was delayed 30 minutes
because of a 30 minute thunderstorm that was going thru chicago that
affected traffic. by the time we finally landed in chicago ... flights
were experiencing 4hr delays ... because of a 30min reduction in
traffic/thruput first thing in the morning. infrastructure had almost
no resilliancy for handling glitches; instead of the effects of a
glitch being mitigated over the course of the day ... the effect was
actually magnified.

of course, almost any complex system can be organized such that it
gets into negative feedback with the effects of small glitches
magnified as time passes (and frequently having to resort to complete
system shutdown before things recover).

nominally you try and design complex systems to degrade gracefully
under adverse conditions ... as well as graceful recovery.

i tried to do graceful degradation with dynamic adaptive resource
management when i was undergraduate ... it was frequently called fair
share scheduler ... since the default policy was fair share.
http://www.garlic.com/~lynn/subtopic.html#fairshare

trivia ... while 6/23/69 unbundling announcement started the pricing
of application software ... kernel software was still free. the (later
re)release of the resource manager represented the first transition to
kernel software pricing
http://www.garlic.com/~lynn/subtopic.html#unbundle

when we were doing the original payment gateway
http://www.garlic.com/~lynn/aadsm5.htm#asrn2
http://www.garlic.com/~lynn/aadsm5.htm#asrn3

we were looking at trying to translate payment transactions from a
circuit based infrastructure where you could get SLA (service level
agreement) contracts (with things like financial penalties for not
meeting contracted service levels) to internet environment. the
original cut had just been to to remap the message formats from
circuit based infrastructure to packet based infrastructure that
carried with it none of the normal service level processes. we were
looking at formulating compensating procedures for internet operation
to try and approximate circuit-based service guaranetees.

In the middle of this period the internet transitioned to hierarchical
routing ... in part because of the large growth. what was left was
getting redundant connections into multiple points in internet
backbone ... and using multiple a-records to map all the different
ip-addresses to the same payment gateway.

previously, you could have had multiple different attachments to
different places in the internet backbone ... and be able to advertize
lots of diverse routings to the same ip-address. with the change to
hierarchical routing, the payment gateway had to have lots of
different ip-addresses to that the diverse paths could be correctly
routed.

misc. recent postings about trying to do business critical data processing
on the internet
http://www.garlic.com/~lynn/2005.html#18 IBM, UNIVAC/SPERRY, BURROUGHS, and friends. Compare?
http://www.garlic.com/~lynn/2005b.html#40 [Lit.] Buffer overruns
http://www.garlic.com/~lynn/2005d.html#42 Thou shalt have no other gods before the ANSI C standard
http://www.garlic.com/~lynn/2005f.html#38 Where should the type information be: in tags and descriptors
http://www.garlic.com/~lynn/2005h.html#16 Today's mainframe--anything to new?
http://www.garlic.com/~lynn/2005i.html#42 Development as Configuration
http://www.garlic.com/~lynn/2005k.html#7 Firefox Lite, Mozilla Lite, Thunderbird Lite -- where to find
http://www.garlic.com/~lynn/2005k.html#17 More on garbage collection
http://www.garlic.com/~lynn/2005n.html#30 Data communications over telegraph circuits

for some routing topic-drift ... in the past i got the opportunity to
rewrite (reservation system) routes application ...  getting a copy of
the complete OAG for all (commerically scheduled) flight segments
world-wide. random refs:
http://www.garlic.com/~lynn/96.html#29 Mainframes & Unix
http://www.garlic.com/~lynn/2000f.html#20 Competitors to SABRE?
http://www.garlic.com/~lynn/2001d.html#69 Block oriented I/O over IP
http://www.garlic.com/~lynn/2001d.html#74 Pentium 4 Prefetch engine?
http://www.garlic.com/~lynn/2002g.html#2 Computers in Science Fiction
http://www.garlic.com/~lynn/2002j.html#83 Summary: Robots of Doom
http://www.garlic.com/~lynn/2003o.html#17 Rationale for Supercomputers
http://www.garlic.com/~lynn/2004b.html#6 Mainframe not a good architecture for interactive workloads
http://www.garlic.com/~lynn/2004o.html#23 Demo: Things in Hierarchies (w/o RM/SQL)
http://www.garlic.com/~lynn/2004q.html#85 The TransRelational Model: Performance Concerns

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

auto reIPL

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: auto reIPL
Newsgroups: bit.listserv.ibm-main,bit.listserv.vmesa-l
Date: Mon, 22 Aug 2005 20:26:31 -0600

Ted MacNEIL writes:

UNIX has had the function for years, since quick re-boots have been
the priority, rather than fewer.

cp67 got fast reboots ... vm370 inherited it from cp67 ...

following has a story about somebody at MIT modifying cp67 resulting
in cp67 crashes 27 times in one day. there is some comment about that
not being possible with multics because it took multics so long to
reboot.
http://www.multicians.org/thvv/360-67.html

i had written tty/ascii terminal support for cp67 while an undergraduate
at the university. i had done some one byte arithmatic for calculating
input line length (couldn't get long lines with teletype).

i think the story is that there was some ascii device at harvard that
had something like 1200-2000 character line length ... that they
needed to connect to the cp67 at mit. the mit system modified various
fields to increase the maximum line length ... but didn't catch the
one byte arithmatic.

some drift ... in the process of adding tty/ascii support .. i tried
to extend the automatic terminal recognition from 1052 & 2741 to
include tty. in theory 2702 terminal controller would allow it.
initial testing sort of worked ... you could dynamically change
the line scanner on each port ... and correctly figure out whether
1052, 2741, or tty and get the correct line scanner. however,
2702 had a short cut where the line-speed oscillator was hardwired
to each port (allowing the correct line-scanner to be dynamically
be set on a port by port basis ... but it wouldn't also change
the port line-speed).

this sort of led to the university having a project to build
a clone controller ... that could support both dynamic terminal
type as well as dynamic line speed. somewhere there is a write-up
blaming four of us for helping spawn the ibm plug-compatible
clone controller business.
http://www.garlic.com/~lynn/subtopic.html#360pcm

on the other hand ... when i was doing getting ready to release
the resource manager .... minor trivia drive ... resource manager
was guinea pig for first priced kernel software ... past postings
on unbundling and software pricing
http://www.garlic.com/~lynn/subtopic.html#unbundle

we did some automated benchmarking process ... which included
extremely severe stress testing
http://www.garlic.com/~lynn/subtopic.html#bench

and there were numerous things that could consistently kill vm370.  as
a result ... before releasing resource manager for vm370, there were
serveral parts of vm370 that i had to redo ... primarily the task
serialization to eliminate all sorts of timing related failures
(as well as all cases of hung and/or zombie users).

not too long after releasing the resource manager ... i got the
opportunity to do some stuff with the disk enginnering and
product test labs in bldg. 14 & 15.

the labs had quite a few "testcells" ... engineering development
hardware that required all sorts of reqression testing. they were
running these test with stand alone processing and dedicated
scheduled processor time. they had tried doing it under operating
system control ... but found that the MTBF (at the time) for
MVS was on the order of 15 minutes with a single testcell.

i undertook to rewrite the i/o supervisor so things would never
fail ... they were eventually not only able to do single testcell
operation in operating system environment ... but eventually
able to do multiple concurrent testcell testing on the same
machine. misc. past posts of disk engineering lab.
http://www.garlic.com/~lynn/subtopic.html#disk

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

How good is TEA, REALLY?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: How good is TEA, REALLY?
Newsgroups: sci.crypt
Date: Tue, 23 Aug 2005 07:57:47 -0600

Paul Rubin <http://phr.cx@NOSPAM.invalid> writes:

If you mean bulk financial data (like a credit card issuer's
database), the institutions that manage that kind of data are very
big into following standards.  Your choices are 3DES and AES.  Even
if Cipher XYZ is almost certainly better, it's not an option.  Maybe
AES has 0.001% chance of being broken and cipher XYZ has only
0.0001% chance, but you face much worse headache if you did
something nonstandard and got unlucky ("why did you think you were
smarter than all those experts?") than if you followed the standard
and the experts turned out to be wrong.  So, for that kind of
application, whether TEA (or CAST, or Serpent, or RC4, or whatever)
is technically good or not is absolutely irrelevant.  You have to
use 3DES or AES, end of story.

part of the issue isn't even technology

there are financial industry standards, at the ISO level, tc68
http://www.iso.ch/iso/en/stdsdevelopment/tclist/TechnicalCommitteeDetailPage.TechnicalCommitteeDetail?TC=68

and in the US, the ISO chartered organization is X9
http://www.x9.org/

and a couple years ago, nist shifted from requiring that all NIST
standards have to be written from scratch ... to allowing direct
adaptation of X9 standards.

in instances involving legal dispute .. if the institution has
deployed something other than specified in some standard, the burden
of proof can shift to the institution ... otherwise the burden of
proof is on the other party.

there was some case in germany from a couple years ago involving
somebody having lost some reasonably large sum of money and the
financial institution was still using some technology that was no
longer covered by standard. instead of the party having to prove it
was the financial institutions fault ... the burden of proof shifted
to the financial institution to prove that it couldn't have been their
fault.

in that sense, rather than being a technology risk management issue, it
becomes a litigation risk management issue.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Data communications over telegraph circuits

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Data communications over telegraph circuits
Newsgroups: alt.folklore.computers
Date: Tue, 23 Aug 2005 07:35:56 -0600

Morten Reistad writes:

The statues of limitations are 30 years.

It applies to details in your service, and all classified data you may
have touched. Configurations, actual clearances, personell information
etc; and daily orders. For some this is a grey area, as they consider
their words as officers to be binding, but there is no legal recourse
to any sanctions against you after 30 years.

I was honorably discharged on April 2nd 1984, and the 30 years count
from that date.

You are permitted to write essays about what you know and put them in
escrow until the statute of limitations expire. They will inherit the
classifications from the author(s).

we were walking down an aisle at some past supercomputer conference
with a couple people who stopped and said they should go another
direction ... up ahead was a booth from some location that wasn't
supposed to exist (at least within recent history covered by some
regulation or another, which supposedly would require them to report
if they ever heard it mentioned).

and somewhat obscurely related to that location.
http://www.garlic.com/~lynn/95.html#13

and some drift on a topic that is almost related to the
subject line (from comp.protocols.tcp-ip)
http://www.garlic.com/~lynn/2005o.html#24 is a computer like an airport?

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Penn Central RR computer system failure?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Penn Central RR computer system failure?
Newsgroups: alt.folklore.computers,misc.transport.rail.americas
Date: Wed, 24 Aug 2005 13:15:37 -0600

hancock4 writes:

Would anyone be accurately familiar with some of the specifics of the
PC's computer failure?  Did they in fact try to merge two incompatible
systems?  Were system failures really a contributor to the bankruptcy?
How sophisticated were their systems in 1968 (computers generally
weren't that fancy back then like today)?

in 1970 ... B&M commuter in boston area was part of PENN
bankruptcy .... cascading effect of east coast railroad bankruptcies
in that time-frame
http://www.nashuacitystation.com/bostonmaine.html Boston and Main Raiload

one of the things mentioned was that there had been deferred track
maint.  on major east coast routes going back 10 years or more (money
going for other purposes, salaries, dividends, bonuses, etc).

there were railroad ties that you could stick your finger into ...
and speed limits were 10-15 mph (because of poor track condition) on
sections that use to be 80mph.

there was joke about one section called the freight car graveyard out
near acton ...  where the track condition was so bad that speed limit
for freight trains was 5pmh ... and even with that there were frequent
derailments (and pieces of abandoned freight cars laying about).

i got the impression that quite a bit of the PENN system had been run
that way ... lots of money going to places other than the nuts and
bolts of railroad operation.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Penn Central RR computer system failure?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Penn Central RR computer system failure?
Newsgroups: alt.folklore.computers,misc.transport.rail.americas
Date: Wed, 24 Aug 2005 18:34:10 -0600

John McCoy writes:

I don't know any specifics about the computers, but in general
everything that the NYC did one way, the PRR did differently.
Everything involved merging incompatible systems.

i remember reading an article a couple years ago about union pacific
acquisition of southern pacific (1996?).

the article was about railroad gridlock trying to get harvested grain
to houston out onto ships for delivery overseas.

supposedly the port and surrounding area didn't have sufficient
staging area for (big?)  increase in grain shipment traffic ... so
southern pacific people got very cleaver about carefully staging
trains thruout texas to avoid gridlock in houston area.

all of that institutional knowledge and expertise was lost in the
union pacific take-over ... with union pacific attempting to do
business as usual railroad operations; resulting in enormous freight
car traffic jams in the the houston area.

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

auto reIPL

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: auto reIPL
Newsgroups: bit.listserv.ibm-main,bit.listserv.vmesa-l
Date: Thu, 25 Aug 2005 08:11:46 -0600

Anne & Lynn Wheeler writes:

we did some automated benchmarking process ... which included
extremely severe stress testing
http://www.garlic.com/~lynn/subtopic.html#bench

and there were numerous things that could consistently kill vm370.  as
a result ... before releasing resource manager for vm370, there were
serveral parts of vm370 that i had to redo ... primarily the task
serialization to eliminate all sorts of timing related failures
(as well as all cases of hung and/or zombie users).

re:
http://www.garlic.com/~lynn/2005o.html#25 auto reIPL

part of the automated benchmark process was to generated a new kernel,
automated rebooting, automated bringing up a set of emulated virtual
address spaces and having them running a variety of workload scripts.

a master script was built ... so the whole process could be repeated
with a wide variety of different workloads and configurations ... and
run unattended.

part of the benchmarking process was creating the "autolog" command
... which was then picked up and included in the base vm370 release 3.

for the final validation of the resource manager ... there was a set
of 2000 benchmarks run that took three months elapsed time. the first
1000 or so benchmarks were selected from points on the surface of
workload envelope and relative even distribution of points within the
envelope ... with 100 or so workloads that were points way outside the
workload envelope to see how things gracefully degraded.

the workload envelope was sort of built up from years of collected
performance data from a wide variety of internal datacenter
operations. combridge science center
http://www.garlic.com/~lynn/subtopic.html#545tech

had nearly 10 years of round-the-clock performance data. there was
then hundreds of other datacenters that had lesser amount of
information ... but also represented quite a wide variety of
workloads.

there was also an analytical performance model written in APL by the
science center. some customers may have run into it dealing with
salesman. HONE was a vm370-based time-sharing system that provided
world-wide online facilities for all the sales, marketing and field
people. various system and performance models had been developed at
the science center in the early 70s. One of these evolved into the
performance predictor available to sales & marketing on the HONE
system. sales/marketing could acquire approximate workload and
configuration information from the customer and input it into the
performance predictor and ask "what-if" questions (in the 70s there
was quite a bit of APL-based implementations that frequently are done
in spreadsheet technologies these days ... although the APL
calculation capability could be quite a bit more sophisticated).
There were an extensive set of other APL-based applications on the
system for sales & marketing use. For instance, starting with the 370
115/125 ... all equipment orders had to be first run thru a HONE
(apl-based) "configurator".
http://www.garlic.com/~lynn/subtopic.html#hone

In any case, for something like the second thousand benchmarks (as
part of calibrating and verifying the resource manager for release),
the performance predictor was modified to look at the previous
benchmark results and attempt to select interesting operational point
for the next benchmark. Then it would kick-off the seeding of the next
round of benchmark scripts, possibly a kernel rebuild and then
automated reboot/reipl of the system.

note that while cp67 had automated/fast reboot ... it came up as a
relatively stale system. as cp67 (and vm370) evolved, there become
more and more services that had to be manually activated (after
reipl/reboot). The "autolog" command that was originally developed for
automated benchmarking was adopted as part of production reipl/reboot
to automatically activate the various system services that were being
created. This also kicked off the basis of automated operator
infrastructure.

for some topic drift ... some number of posts about commercial
time-sharing systems starting with cp67 and later migrated to vm370
http://www.garlic.com/~lynn/subtopic.html#timeshare

one of the early cp67 features for 7x24 commercial time-sharing
systems (besides fast, automated reipl) ... was the use of the
"prepare" command for terminal I/O.

typical 360 machines were leased and the datacenter paid based on the
amount of time that was recorded by the processor meter. The processor
meter ran anytime the CPU was running or anytime there was "active"
I/O. early introduction of commercial time-sharing systems could be
expensive for 2nd and 3rd shift operation ... since the active use of
the system could be less than the lease charge based on the processor
meter. changing to the "prepare" ccw I/O command for terminal I/O
allowed the terminal I/O to suspend from the channel until the next
character (and therefor if nothing else was going on, the processor
meter would stop).

the change for the "prepare" command probably seems trivial these days
... but back then, it could represent a significant economic issue for
a commercial time-sharing operation trying to offer 7x24 service.

for some other topic drift (somewhat availability/ras related)
... several years ago we were talking to one of the large financial
transaction network operation. they had claimed that the two primary
things responsible for their one hundred percent availability over the
previous six-plus years were:

• ims hot-standby
• automated operator

I think there was a study in the early 80s that showed that primary
cause of system outages had shifted from hardware related to
non-hardware related (software and human "mistakes").

my wife had done her stint in POK in charge of loosely-coupled
architecture where she created peer-coupled shared data architecture.
except for the ims hot-standby people ... it didn't see much uptake
until parallel sysplex. misc. past peer-coupled shared data
architecture.
http://www.garlic.com/~lynn/subtopic.html#shareddata

although, one might claim that some of that was leveraged when
we did the ha/cmp product
http://www.garlic.com/~lynn/subtopic.html#hacmp

--
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/

Is symmetric key distribution equivalent to symmetric key generation?

From: Anne & Lynn Wheeler <lynn@garlic.com>
Subject: Re: Is symmetric key distribution equivalent to symmetric key generation?
Newsgroups: sci.crypt
Date: Thu, 25 Aug 2005 10:29:25 -0600

daw@taverner.cs.berkeley.edu (David Wagner) writes:

Well, that's a different question!  There are many possible ways.
You might have that person's public key.  You might have a secure
channel through which you can request the public key.  You might be
able to ask some other trusted party ("Verisign") for that person's
public key.  That person might provide a certificate signed by some
other trusted party ("Verisign").  One might use a hierarchical PKI
(a la x509) or a web of trust (a la PGP).  There are many answers,
and which one is right is going to depend upon the application, upon
the trust model, upon compatibility requirements, etc..

But whether you use a key transport protocol, or a key agreement
protocol, doesn't change the need to answer the question, and
doesn't change the set of answers available.  How to get the other
endpoint's public key is orthogonal to what kind of key exchange
protocol you use.

note ... in general ... all of the public key related infrastructures
have some out-of-band trust protocol.

basically the dependent or relying party ... has a trusted repository
of (trusted) public keys that was loaded by some out-of-band trust
process.

one can consider that originally trusted public key repositories
consisted only of indivi