|
|
| OSI and TCP: A History, by Peter H. Salus |
 |
|
Sunday, February 20 2005 @ 12:58 PM EST
|
When we reported the awarding of the Turing Award to Vint Cerf and Robert Kahn for the TCP/IP protocol, we had a lot of Unix greybeards commenting and reminiscing. Unix historian Peter H. Salus left the following comment: "It's important to recognize, while we laud Vint and
Bob, that *all* the work on the ARPANET was open and
unencumbered by patents, etc.
"Steve Crocker, Ray Thomlinson (mail), Dave Walden,
Len Kleinrock, Jon Postel, and nearly 100 others
get credit for these wonders, which I'd not even
thought about before the mid-1970s.
"There was a decades-long battle between OSI and TCP/IP,
with all the manufacturers and the PTTs lined up
against TCP/IP. TCP/IP won because it worked, because
it had been implemented, and because the networkers
actually used it.
"Bouquets to Bob and Vint are well-deserved. For over
three decades of work each."
We also then had a great suggestion from iceworm, seconded by several, including me, that Salus, who is a Groklaw member I am happy to say, might like to write a more complete article for Groklaw, to fill in some of the blanks in the story from the New York Times.
Iceworm wrote: "Perhaps it would be well to unravel the 'OSI' stack and why it did not become the standard for the Internet. I dimly remember discussing it in one class meeting (well, part of the class) in the semester I took a course in networking about five years ago. I remember (dimly) some comments along this same line of why TCP/IP won the day, but I don't remember the details. "We did use the "OSI" concept to make short presentations. We never did examine the differences between the "OSI" stack and the TCP/IP concept. Again, perhaps there is someone who could describe these differences in layman's language which might be of interest to the Groklaw community.
"Finally, I read a little on the concept behind IBM's OS/2. There seemed to be some bit of the "OSI" concept here also in that there were "application" and "display" layers, if I remember correctly. So a further question comes to mind, namely, "Did OS/2 become a niche player because it was closed and could not compete with the monopoly OS, or was its underlying design deficient as compared to the Unix kernel design?" Peter graciously agreed to write an article giving us the history, and here it is. I love the barfbag detail. For the detailed comparison of stacks
that iceworm wanted, those are found in Salus' book, "Casting the Net: From ARPANET to INTERNET and Beyond
". Please feel free to add your own memories here, so it will be a permanent record for our enjoyment and for historians in the future. Some are reportedly thinking of writing a letter of formal complaint to the Association for Computing Machinery (ACM), which gives out the Turing Award, because of perceived issues with MCI, Cerf's employer, and spam, protesting the award to Mr. Cerf. I hope they don't embarrass themselves with such an ungrateful, misguided, and unkind gesture. Spam is awful, but it is not, I must say, the only awful thing people can do to one another. Internet thuggery is plenty awful too. Mr. Cerf should be allowed to enjoy this wonderful honor, which is well-deserved, in peace. We ought to all be allowed to disagree with each other on issues of concern to us all, without being attacked and painted in evil brush strokes. And there truly is no connection between this honor and spam. It's just not nice to ruin this moment, and I hope they will reflect and desist.
**********************************
OSI and TCP: A History
by Peter H. Salus
[The following is an abbreviated conflation of pp. 117-126 of my
Casting the Net (1995) and my review of Padlipsky, The Elements
of Networking Style (1985; 2000) to appear in //www.unixreview.com.]
In 1977, the British Standards Institute proposed to ISO, the
International Standards Organization, that a standard architecture
was needed to define the communications infrastructure. (As I've
said and written repeatedly, this, as with IFIP, CCITT, and other
efforts, shows how the road to hell is paved with good intentions.
Because X.25 was unsatisfactory, the IFIP Working Group was set up
in the vain hope that the technical community could block the political
arena that is ISO. It didn't.)
ISO set up a subcommittee of a technical committee to study this
(ISO/TC 97/SC 16). The next year (1978), ISO published its
"Provisional Model of Open Systems Architecture" (ISO/TC 97/SC 16 N 34).
This was labelled a "Reference Model," and referred to as OSIRM (Padlipsky,
my favorite curmudgeon, labelled it ISORM -- "eye-sorm").
In general, ISORM was based on work done by Mike Canepa's group at
Honeywell Information Systems, which came up with a seven-layered
architecture, which itself owed a lot to IBM's proprietary Systems
Network Architecture (SNA). SNA had been announced in 1974 and its
seven layers do not correspond exactly to ISORM's.
TC 97/SC 16 turned over proposal development to ANSI, the American
National Standards Institute, to which Canepa and his technical
lead, Charlie Bachman, presented their layered model. This, in turn,
was the only proposal presented to the ISO subcommittee in March 1978.
It was accepted and published immediately.
A "refined" version of the ANSI submission to ISO appeared in June
1979. This published version is (gosh!) nearly identical to that of
Honeywell in 1977.
Dave Walden and Alex McKenzie had pointed out that both virtual
circuit and datagram services were valuable. "An international
standard would do well to support both," they wrote. The 1977-1979
models were such that the extant host-host protocols didn't fit.
To me, there was a clear opposition: ISO was trying to construct
a nice set of geometric figures that would be a "tidy model"; the
ARPANET folks were interested in getting things to actually work --
they wanted to push bits around the system.
The irascible Padlipsky has described the OSI model as a pair of
high-rises with parking garages. The basic model had a pair of
seven-storey buildings; reality complicated things.
John Quarterman pointed out a decade ago that "OSI specified before
implementation. So specification took forever and implementation
never happened. ... OSI and IP started about the same time (1977).
OSI wandered off into the weeds and got lost. IP won the race."
There are two more things to point out.
The EC (now the EU), the European and the Japanese
PTTs all supported ISORM. They pushed hard for it in a variety
of ways. But, by the end of the 1970s, TCP/IP was a suite with a
real installed base. Those implementations had evolved between
1974 and 1978 and had become accepted by the technical community.
By supporting a theoretical specification with no implementation,
the Europeans and the Japanese became identified with the "old"
telephony and telegraphy.
As all the ARPANET work had been supported by the US DoD, the
EC and Japan appeared to mired in NIH -- not invented here. And
the European and Asian governments feared that this would give
US manufacturers a boost.
Finally, as the ARPANET (soon to become the Internet) had been
governmentally funded, it was (as we would now say) freely
distributable.
But real money was involved. Future profits were at stake.
The result was that many companies wasted vast sums of cash for a decade
trying to produce OSI products while the networking community
moved ahead with TCP/IP.
As early as 1986, Jack Haverty and Gary Tauss were sure that the
market had "the choice between a well understood, widely implemented
protocol set, namely the Internet Suite, and an evolving, incomplete,
not widely available set, representing the OSI model" (Government
Data Systems 15.3, April/May 1986).
The fact that local area networks and the manufacturers of desktop
equipment supported TCP/IP meant that there was an ever-waxing
infrastructure built on TCP/IP in Europe. The changing situation
in Eastern Europe and in the (then) USSR, meant ever-increasing
opportunities for TCP/IP, as all the protocols were in the public
domain and thus no fees were payable.
By a dozen years ago, while there were some OSI networks, there was
no growth. Store-and-Forward networks like Bitnet were shrinking,
and the growth of Fidonet and UUCP had slowed dramatically.
The US DoD played no small role in proliferating confusion. I don't
want to go into either AUTODIN (Defense Information Network) nor
AUTODIN II here. But, in its customary fashion, the Department let
contracts to all the usual suspects: Western Union, Ford Aerospace,
CSC, and Mitre all received contracts. "Western Union knew how to
talk to the military," Dave Walden told me over a decade ago. "We
didn't. For example, they wanted security. We [BBN] said: 'This
is easy. This is hard.' Western Union said: 'Of course we can do
that.' So they got the contract."
Padlipsky, I was told on good authority, once marched into a
meeting with the vendors, pulled an airline "barfbag" from his
attache case, put it on the table and said "OK, you guys, I'm
ready for you."
It turned out that Western Union couldn't do the job. The DoD
finally convened an "execution squad" and killed the project. It
had taken over 15 years and untold millions of dollars, but finally
Vint Cerf, Bob Kahn and others persuaded the military brass that
the ARPANET/Internet protocols reliable, available and survivable.
I've been told that it was Frank Carlucci himself, when he was
Secretary of Defense (1987-1989), who actually signed off on
terminating the Western Union agreement.
Over the decades, the National Bureau of Standards (now the National
Institute of Standards and Technology) issued Federal Information
Processing Standards, which set down the requirements for vendors
interested in purveying to the government (an enormous market).
In the late 1970s, a FIPS was issued requiring OSI capability. So
manufacturers dutifully supplied buttonholes and velcro, should
anyone ever actually implement ISORM. Only in 1994 was the FIPS
finally rescinded.
I think this is relevant and important because, though it took
over 15 years, technology won. The "good guys," led by Cerf, Kahn,
Walden, and others defeated the governments, PTTs, and giant
corporations by using well-engineered, open protocols -- stuff
that worked and was robust.
Copyright © 2005 Peter H. Salus
|
|
|
|
| Authored by: Nick on Sunday, February 20 2005 @ 01:12 PM EST |
And one I saw is referring to Mr. Cerf in a couple of places in the
introductions as Mr. Cert.[ Reply to This | # ]
|
| |
| Authored by: DBLR on Sunday, February 20 2005 @ 01:24 PM EST |
Please place your Off Topic post and links here.
Charles
---
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is
a well-armed lamb contesting the vote."
Benjamin Franklin.
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 01:42 PM EST |
Hehe, if tcp is stable, i wonder what osi looked like.
All i hear now that the internet protocols need a revamp.
[ Reply to This | # ]
|
| |
| Authored by: artp on Sunday, February 20 2005 @ 01:47 PM EST |
Imagine my surprise when I ran into an OSI stack on the telecommunications
network inside a hardened concrete bunker that was a CLEC. I was bringing up the
hardware in preparation for installation of some Alcatel software that worked
with the network monitoring software. This would gave been about 1999.
Before that, I had heard about OSI, but had never seen one. Is it still used
anywhere else?[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 01:47 PM EST |
Does anyone know the OS/2 story?
[ Reply to This | # ]
|
- So-Did OS/2 become a niche player because it was closed? - Authored by: Anonymous on Sunday, February 20 2005 @ 01:54 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: emmenjay on Sunday, February 20 2005 @ 01:57 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: Anonymous on Sunday, February 20 2005 @ 02:25 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: Anonymous on Sunday, February 20 2005 @ 03:22 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: OnlyAnEgg on Sunday, February 20 2005 @ 03:35 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: T O'B on Sunday, February 20 2005 @ 03:39 PM EST
- Everybody gets an opinion - Authored by: TomWiles on Sunday, February 20 2005 @ 05:11 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: garbage on Sunday, February 20 2005 @ 07:28 PM EST
- How did Ed Iacobucci's defection from IBM, with 14 Developers, to create Citrix, affect OS/2? - Authored by: Anonymous on Sunday, February 20 2005 @ 08:01 PM EST
- OS/2 as I recall - Authored by: Anonymous on Sunday, February 20 2005 @ 09:22 PM EST
- OS/2 was a victim of anti-competitive practices - Authored by: grouch on Sunday, February 20 2005 @ 11:49 PM EST
- So-Did OS/2 become a niche player because it was closed? - Authored by: elderlycynic on Monday, February 21 2005 @ 05:35 AM EST
- Pay for Two Windows licenses.... - Authored by: NetArch on Monday, February 21 2005 @ 12:58 PM EST
| |
| Authored by: emmenjay on Sunday, February 20 2005 @ 01:52 PM EST |
As late as 1989, I remember being taught that OSI would eventually be fully
implemented and adopted, and that TCP/IP was (in the long term) doomed. I
remember hoping that my lecturer was wrong, as OSI was extraordinarily complex
and TCP/IP seemed much easier to understand.
Of course, at that stage TCP/IP ruled the Unix world (and the internet was very
Unix-centred), but as networks became popular in offices it was Novell's IPX/SPX
that ruled the roost. Fortunately, that has just about gone with the dodo's.
[ Reply to This | # ]
|
| |
| Authored by: archonix on Sunday, February 20 2005 @ 01:54 PM EST |
Fascinating. :)
I've just been playing with IPV6 on my new Gentoo system (64 bit, pukka putah!)
with varying results. It's all so fascinating to me because I only understand
the ideas on a very basic level.
When I was in college (British college, not american) I was taught the OSI model
of networking in all but name and IP was glossed over, explained in the context
of an OSI stack. This was in the early 90s. I suspect the tutor had some sort of
grudge against the IP way of things...
---
18th June 2005: The day my freedom is voluntarily handed over to a woman in
exchange for a small ring. And it doesn't even make me invisible...
Graham Dawson[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 01:59 PM EST |
If I recall correctly, one of the major European (and Japanese, Chinese etc?)
concerns with the TCP/IP protocol was that it does not adequatly deal with
characters not used in the English language, whereas OSI would be
compatible. The English speaking community may not fully appreciate this
concern. But just consider the poor guy Östen Ås (spellt Oesten Aas for those
of you that cannot correctly visualise diacritical characters with your
browsers): in e.g Sweden his e-mail address wound be Osten.As@site.se.
Well, "osten as" literally means " the guy who smells like a well
putrified
carcass". This type of incompatibility between "computer-speak"
and the
actual languages computers are thought to serve is a major international
concern that unfortunately the technology leading but monolinguistic
American programmers usually fail to comprehend.
Therefore, I do not quite appreciate Peter Salus' comments on technologically
oldfashioned Europeans, in particular not after the previous comments by the
US government about an "old Europe". In the end, Europe may end up
with a
better law governing software patents, and the continent is far more
advanced in the use of mobile telephony as compared to the US.
IANAG (I am not a geek), so I might be wrong on these technical matters. If
so, I apologise.
bjolf[ Reply to This | # ]
|
- OSI and TCP: A History, by Peter H. Salus - Authored by: Pug Majere on Sunday, February 20 2005 @ 02:26 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Philip Stephens on Sunday, February 20 2005 @ 02:28 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Anonymous on Sunday, February 20 2005 @ 02:38 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Anonymous on Sunday, February 20 2005 @ 02:38 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: John Hasler on Sunday, February 20 2005 @ 02:40 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Anonymous on Sunday, February 20 2005 @ 02:40 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Anonymous on Sunday, February 20 2005 @ 02:48 PM EST
- Osten - Authored by: johan on Sunday, February 20 2005 @ 11:02 PM EST
- Troll, Joke, or just uninformed - Authored by: Anonymous on Sunday, February 20 2005 @ 11:45 PM EST
- OSI and TCP: A History, by Peter H. Salus - Authored by: Anonymous on Monday, February 21 2005 @ 08:42 AM EST
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 02:00 PM EST |
| I lived thru all this near the beginning of my first years
out of college,
even implemented bits of stuff on both
sides of the fence (and there were many
sides to this
fence lots of other proprietary architectures - remember
BNA?,
DNA?, etc even Appletalk that's still with us)
The big issue IMHO was
really datagrams vs. virtual
circuits - the last largely driven by the fact
that large
phone companies were involved on the OSI side (it was the
only way
they knew how to charge for phone service, and
they they had a culture of
reliability - think '911').
Building good working networking systems is
HARD - people
forget that TCP was still being tweaked decades later.
Anyway this discusion leaves out one thing that I found
quite ugly about the
whole thing was the whole US vs. the
rest of the world (though mostly Europe)
thing that came
out of it people lined up on both sides largely based on
where
they were living - like the cold war that it
somewhat modelled at the time in
my experience each side
was relatively ignorant of the realities of the other [ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 02:44 PM EST |
Actually, Windows "sold" more than OS/2 primarily because of preloads
and some bundled monopoly practices. (Which handed Microsoft an antitrust court
defeat, after the damage had been done.) Had the market been more open, OS/2
would have done better. (It was widely considered the technically superior
product.) Both were closed source products, however.
By the way, IBM introduced OSI products for OS/2 as early as 1989 and withdrew
them in 1994. But TCP/IP for OS/2 dates to the same era and shipped the same way
(extra-cost option). In other words, IBM offered its customers a fair choice. By
1994, IBM had the first "major" PC operating system (OS/2 Warp) with
built-in TCP/IP. When it first shipped there was a Gopher client, but IBM had a
downloadable Web browser (Web Explorer) available from Day 1, later rolled into
an update (OS/2 Warp Connect) -- another first. (Remember, this was a year
before Microsoft almost missed the whole Internet phenomenon.) Then, in 1996,
IBM shipped the first "major" operating system (OS/2 Warp Version 4)
to include Java. The server releases (OS/2 Warp for e-business) were also very
aggressive in adopting new standards. IBM had one of the first DDNS
implementations, for example.
And these were really technically excellent implementations. IBM's
implementations were "BSD-like" (BSD code with platform adaptations)
and they didn't take Microsoft's "Not Invented Here" course -- a
course which is still killing their customers with security problems.
Hope that fills out the historical record. IBM was pretty aggressive in shipping
then-nascient Internet standards in its flagship PC operating system.[ Reply to This | # ]
|
| |
| Authored by: arch_dude on Sunday, February 20 2005 @ 02:48 PM EST |
X.25 ws the first commercial response to the ARPANET, and it was a qualified
success. Not counting Telex, the first large public data nets in the US and
world were X.25 nets. This included Telenet, Tymnet, and Compuserve in the US.
Overseas, the X.25 nets were mostly run by the PTTs. The nets and the hosts ran
X.25 while the subscribers used async modems. This was perfectly reasonable
since a synchronous link to the home was prohibitively expensive. The big X.25
nets started in the mid 1970's, were dominant in the mid-80's and remained
viable in the early '90s. The last host was taken off the Telenet (Sprint)
network last month.
In my opinion, TCP/IP won because it was open, not because it is a better
protocol than X.25. More specifically, The IETF won and CCITT lost, because the
IETF organizatonal model (emergent order from chaos) was crushingly superior to
the CCITT model (UN-style top-down, run by and for the benefit of the telecoms
operators.)
To make a change to the X.25 protocol, you needed to be a member of a company
that had representation on your country's standards body (ANSI) who could then
participate on the CCITT comittees. These comittees met periodically to
promulgate and progress drafts which ultimately were published as standards,
once every four years!!! If you wanted to read one of the standards, you had to
pay for a paper copy that you would then receive about 2 weeks later. Those
copies were expensive. If you wanted a reference implementation, you had to pay
a vendor.
By contrast, You can jsut show up at an IETF meeting. An IETF standard is still
called an RFC (Request for Comments) when it is finalized, and you could get one
by asking the author to e-mail it to you. You could almost always get some
reference source code to play with.
X.25 was invented by BBN (which spawned Telenet.) Telenet was founded within BBN
by Larry Roberts. If there was ever a "father of the internet" it was
Larry Roberts. He was the project manager for the ARPA net within DARPA.
As Roberts saw it however, there was no way to commercialize the ARPANET
protocols because there was no was to bill for connections. In the ARPANET
protocols and their decendants, connection state is maintained within the
end-user equipment, at the TCP layer. Roberts also believed that a datagram
protocol is inherently less efficient than a VC protocol. Roberts was correct in
detail in both cases. He did not realize that these details would become
irrelevant.
Roberts' solution to this was to evangelize for an SVC protocol. In essence, his
people (Don Wier, Peggy Karp, John Holmblad, etc.) invented X.25 and Roberts
convinced the PTT community to use it. The PTTs were aghast at the thought of a
protocol that let the user take control of the conections, so they were very
receptive to X.25.
The above opinions come from my tenure at Telenet in 1979-1983, which was after
many of the events I relate here. However, it was the accepted wisdome at the
time.
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 03:36 PM EST |
"There was a decades-long battle between OSI and TCP/IP, with all
the manufacturers and the PTTs lined up against TCP/IP. TCP/IP won because it
worked, because it had been implemented, and because the networkers actually
used it. Sounds a bit like Hurd vs the Linux kernel... Linux won
out because it was here, now and works... whereas, HURD isn't really ready yet,
has been specified to death almost, as RMS was taking forever to get the design
right before coding anything... and now the HURD team are going to change
micro-kernels underneath the whole thing... [ Reply to This | # ]
|
| |
| Authored by: teknomage1 on Sunday, February 20 2005 @ 03:40 PM EST |
This reminds of when I was in high school, taking classes to be a Cisco Network
Engineer. circa 1998. We had to learn the full seven layer OSI stack but it was
presented as a conceptual model rather than a concrete thing. We then had to
learn the 4 layer TCP/IP stack and compare those to their counter parts in the
OSI model and such. I never realized OSI was supposed to be implemented.[ Reply to This | # ]
|
| |
| Authored by: RedBarchetta on Sunday, February 20 2005 @ 04:00 PM EST |
I seem to remember a time when NIC (network interface card) manufacturers
included what was called a "packet driver" with their product. This
was a higher-level interface driver that allowed one to grab and insert raw
ethernet (802.3) packets into the wire. There were also a host of other packet
drivers included, just in case you needed one of 12 different drivers.
Of course, the packet driver had nothing to do with the TCP/IP protocol itself,
but it does illustrate the disarray of pack protocols that existed a decades or
two ago.
SAP or SPX anyone?
---
Collaborative efforts synergise.[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 04:02 PM EST |
"Some are reportedly thinking of writing a letter of formal complaint to
the Association for Computing Machinery (ACM), which gives out the Turing Award,
because of perceived issues with MCI, Cerf's employer, and spam, protesting the
award to Mr. Cerf. I hope they don't embarrass themselves with such an
ungrateful, misguided, and unkind gesture. Spam is awful, but it is not, I must
say, the only awful thing people can do to one another. Internet thuggery is
plenty awful too. Mr. Cerf should be allowed to enjoy this wonderful honor,
which is well-deserved, in peace. We ought to all be allowed to disagree with
each other on issues of concern to us all, without being attacked and painted in
evil brush strokes. And there truly is no connection between this honor and
spam. It's just not nice to ruin this moment, and I hope they will reflect and
desist."
...just realize that they feel hopeless and helpless. It is their opportunity
to feel important (by foolishly thinking the can influence public preception and
policy). It is best to just ignore them as I hope Mr. Cerf, and most certainly
ACM will.
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 04:13 PM EST |
| And now you get Linuxes for Windows SuSE-for-W
indows
Interoperate with anything ! Torrents
here
2.4 kernels like Knoppix for
Windows perform best, but they all function just fine. [ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 04:33 PM EST |
In the late 1970s, a FIPS was issued requiring OSI capability. So
manufacturers dutifully supplied buttonholes and velcro, should anyone ever
actually implement ISORM. Only in 1994 was the FIPS finally
rescinded.
I think this is relevant and important
because, though it took over 15 years, technology won. The "good guys," led by
Cerf, Kahn, Walden, and others defeated the governments, PTTs, and giant
corporations by using well-engineered, open protocols -- stuff that worked and
was robust.
I'm not sure that the lesson to take away is that
"using well-engineered, open protocols" is what leads to triumph.
As I recall
the U.S. government also had a standard that said that all computer systems were
to run Unix. Government offices bought MS Windows anyway, and today it's the
de-facto standard in the government as everywhere else. (Take a look at the Security Space
webserver market share reports for the .gov and .mil domains. This is the
only market that the Microsoft IIS web server has any traction at all
compared to Apache.)
No, I'd say the true lesson lurks somewhere
amoungst "people will buy something that works well enough", or
"decentralized
is often more productive than centralized".
Or maybe "it's easier to sell lots
of cheap items than fewer
more expensive items."
Microsoft Windows is
what ran the desktop PC, and people want control over their computing
environment so government
offices, and offices in general -- software
developers
included,
bought PCs instead of subjecting themselves to the
controls
and restrictions of the Unix administrators.
The offices could come up with the
money because PCs were
cheap. Unix boxes required a budget.
Karl O.
Pinc <kop@meme.com>
[ Reply to This | # ]
|
| |
| Authored by: ghbyrkit on Sunday, February 20 2005 @ 05:42 PM EST |
During 1986 thru 1988 or so, I worked for Ungermann-Bass in Michigan on the
MAP/TOP initiative targeted for manufacturers like GM and Alcoa who wanted to go
the 'OSI Model' route. The physical network was broadband ethernet, which, like
cable TV, allowed multiple channels of ethernet on the same cable.
I worked on the lower 4 layers (Transport on down to the wire), taking over from
various predecessors and finishing their work. Others in the group had
implemented the upper 3 layers, and it worked. The problems were mostly with a
component at the top, ASN.1 (Abstract Syntax Notation version 1), which exists
and is used to this day, but outside the OSI model/framework. ASN.1 was
designed to deal with providing a representation of integers, floats, strings
and other data that was valid on both source and destination platforms. It
handled things like bit-endianness and language of strings.
IIRC, it was ASN.1 that was never quite done or working at that time, with
changes and drafts proposed to deal with the issues that implementation revealed
(bugs or lack of specificity and/or clarity in the spec).
So while we had a working 7 layer stack, one actually needed applications
written that used that model, and they never really existed by the time the US
government decided to go with its 'bird in the hand' TCP/IP, rather than choose
to follow the words of the FIPS that would have otherwise made the OSI stack a
mandatory choice.
Was the US government right or wrong? Neither, they were just being pragmatic
and choosing something that already worked. And wasn't over-specified. History
bears out that the choice they made has worked so far.
I've never had cause to use much of the knowledge that I acquired during that
project, but it was a wonderful education in both protocols and how protocols
are designed, implemented and accepted.[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 05:48 PM EST |
| Some reminiscences from one of the shock troops in the TCP vs. OSI vs. SNA vs.
"we-don't-need-no-stinkin'-networking" protocol wars of the mid-Eighties:
[I
was an SNA guru: my team spent years working to distinguish between the SNA
architecture and IBM's occasionally baroque implementations of
that architecture so that our implementation could interwork with theirs.]
The UNIX developers of that era, working down the hall from my group to
port the latest SVRx changes to our hardware, considered that well-implemented
high-speed asynchronous device support was all the networking that UNIX would
ever need. The concept of layered protocols, and the separation of presentation
from transport, was considered a waste of good CPU cycles (which were precious
in that day and age!). A good device driver was all that UNIX would ever need
... Of course, the job of producing device drivers for each of the "intelligent"
terminals of that era (each with a different set of formatting sequences)
turned into an ever-receding race.
SNA for UNIX (which our team was
implementing) was based on the premise that host-to-host interworking would be
the key to the future. We got the principle right, we just got killed by the
details.
Another group spent ~$10e7 to acquire and port an OSI stack,
X.400 mailer and X.500 directory services. That expenditure let us complete in
goverment RFP's where OSI was a checkoff item. We didn't win. We spent
~$10e8 over a decade to implement an SNA stack and services for UNIX that
would interwork with IBM's mainframe SNA ("VTAM"). Guardians of mainframe
computer centers of that era responded to concerns that connecting SNA on UNIX
to VTAM would violate their warranty by rejecting our sales overtures. The TCP
group downloaded the BSD sources, ported them to our system, hired a consultant
or two, and considered the job done. Total expenditure ~$10e6. With no
requirement for expensive software stacks or for teams to understand arcane
interfaces, their efforts slid under the budget radar.
Since the BSD stack
usually had to interwork with another port of the BSD stack, host-to-host TCP
usually came up in hours. Grooming our stack to behave well with a random
service level of VTAM required weeks of painful analysis to determine if the
problem was implementation or software service level. I don't believe we ever
did have to address the problem of OSI stacks interworking (whew!). [ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 06:01 PM EST |
Peter, thank you for your insights. They really help recall the history.
There are a couple of additional things that come to mind that may also help
place the tcp/ip story in context.
First, it is worth recalling that at the time in question, IBM was more or less
finished destroying the last of its meaningful competition in the world of large
machines. The Europeans, hopeful as ever to avoid losing all control of their
computing future and national computing companies, pushed the ISO stack as a
means to prevent IBM and IBM's SNA (Systems Network Architecture, a layered
stack whose basic principles were closely followed by the OSI model) from
becoming the defacto standard for all computer networking and communications.
Their answer was the OSI model made a mandate by ISO ratification. The
European computing community and governments hoped that by mandating support for
a published standard, they could limit the damage from the ubiquity and closed
nature of SNA.
Second, one should recall just how thoroughly iBM came to dominate computing
toward the end of the 1970's and the beginning of the 1980's. Many, many large
customers made SNA interoperability a requirement for the purchase of any
"foreign" (non-IBM) equipment. However, the SNA standards were
whatever IBM said they were and were subject to change at the pleasure of IBM.
I have literally participated in computer installations where Honeywell, Univac,
and Bull equipment could each talk to the other using SNA but none of them could
talk to all of the IBM equipment present. If this sounds rather like the
problem of Microsoft and the SMB protocols as viewed from the standpoint of say
a SAMBA developer, it is.
It was popular among my peers at Honeywell and CDC to refer to the IBM
"department of non-standard standards". It was an early version of
"embrace and extend" where not only the software but also the hardware
could be enlisted in the game of "frustrate the competition". It
began with a parochial technical dialect (say DASD, not disk) and extended to
making marketing features from technically inferior implementations "multi
processing with a fixed number of tasks..."
All by way of saying that it was not purely a matter of obdurate bureaucracy,
stodgy academics, and unimaginative governments that drove the ISO/OSI efforts.
It was also a matter of a perceived open standard alternative and survival. An
illustration, perhaps, that more than "open-ness" is needed to be
successful. The market failure of "InfiniBand" recently demonstrates
that we still have some things to learn.
cheers,
greybeard, who does not have his login set up on this system
[ Reply to This | # ]
|
| |
| Authored by: alextangent on Sunday, February 20 2005 @ 06:16 PM EST |
| See here for a
history of the internet; and specifically here for the
reminsicences of (quoted from the last link) Barry M. Leiner, Vinton G. Cerf,
David D. Clark, Robert E. Kahn, Leonard Kleinrock, Daniel C. Lynch, Jon Postel,
Larry G. Roberts, Stephen Wolff.
--
An Interested Bystander [ Reply to This | # ]
|
| |
| Authored by: ENOTTY on Sunday, February 20 2005 @ 07:37 PM EST |
The OSI seven-layer stack does not appear to have been invented
by the followers of Zen. A cornerstone of Zen is the focusing on
reality and the avoidance of misleading conceptual abstractions.
-Alex Gillespie "Access networks: technology and V5 interfacing"
There's another very funny one I can't quite remember regarding an ISO committee
choosing a wine for lunch and hell.
[ Reply to This | # ]
|
| |
| Authored by: blacklight on Sunday, February 20 2005 @ 10:30 PM EST |
"OSI specified before implementation. So specification took forever and
implementation never happened. ... OSI and IP started about the same time
(1977). OSI wandered off into the weeds and got lost. IP won the race"
It's not a race when the competitor is running off cross-country while you are
doing the 100 meter dash to the finish line.
I was never able to make any practical sense of the OSI model, except as a
baseline frame of reference for comparing network traffic protocols. I
understand the Federal government conned some major vendors such as DEC into
spending (wasting) hundreds of millions of dollars developing OSI model
compliant protocols, all of which suffered from implementation drawbacks such as
lousy performance. It got to the point where vendors finally went into passive
resistance and refused to deal with the Federal government's mandate that the
protocols be OSI compliant. The Federal government, in order to not isolate
itself from the computer marketplace, issued more and more exceptions to its
OSI-only mandate to the point where it simply decided one day to mandate
TCP/IP.
While some good ideas and concepts such as DAP(*) came out of the OSI approach,
my personal opinion is that OSI itself is a piece of crap.
(*) Typical of the OSI bureaucratic approach, DAP was top heavy, over-specified
and a performance killer. A lighter version universally known as LDAP was
implemented instead in the real world.[ Reply to This | # ]
|
| |
| Authored by: blacklight on Sunday, February 20 2005 @ 10:42 PM EST |
Back as in 1996, I would tell interviewers that I had absolutely no desire to
waste my time working for an employer who was too stupid to implement TCP/IP.[ Reply to This | # ]
|
| |
| Authored by: Totosplatz on Sunday, February 20 2005 @ 11:17 PM EST |
I repeat this notice about "Dream Machine - J. C. R. Licklider and the
Revolution That Made Computing Personal", by M. Mitchell Waldrip.
This
is an excellent read, I recommend. Very strong especially on developments in the
fifties and sixties, and strong on the groundwork and foundations of the
Internet. And a very good read. ---
All the best to one and all. [ Reply to This | # ]
|
| |
| Authored by: fgoldstein on Sunday, February 20 2005 @ 11:30 PM EST |
| It's an interesting story that Salus tells, but it's incredibly one-sided. He
knows the origins of the OSI Programme, but doesn't tell what happened on the
OSI side of the TCP/OSI divide. Instead it's a TCP/IP partisan's view of good
guys vs. bad guys, ignoring half of the OSI world.
I have some inside
knowledge, since I worked at DEC when we were spending a huge heap of money on
OSI, much of doing standards work. I'm friends with the guy who was Rapporteur
of the OSI Reference Model for most of its existence, and have heard his side of
the tale. DEC wasted the effort, of course. I also spent a lot of time at BBN.
But it's a complex tale.
As noted by Salus, there was a need for both virual
circuits and datagrammes. However, the real issue was that there were two
entirely different factions both trying to wear the OSI badge. IBM and the PTTs
were both in the virual circuit (connection-oriented) camp, because both were
focused on terminal to host networks, and in that application, virtual
circuits work better. Really. The DEC and ARPAnet factions favored datagrammes
(connectioness), because they were focused on host to host networks, and
in that application, datagrammes work better.
The Reference Model was never
designed to describe implementations. It was designed to divide the work among
subcommittees. The model had a profound error in it; layers 5 and 6 did not fit
below the application layer, but were more properly modeled as optional
application-layer functions. Combine the mistake in the model with the idea
that each layer should be implemented as a module of code and you end up with a
stinking mess! Marshall Rose wrote a successful text, The Open Book,
describing OSI; he also gave away free code (late 1980s) that implmemented each
layer individually. Application atop a discrete Presentation atop a discrete
Session. Conformant, perhaps, but gawd-awful in practice. However, it was the
OSI equivalent of Berkeley's free TCP stack, and it got implemented. We used it
in some products at DEC. A few packets per second, maybe, of throughput.
Managers couldn't resist the price of gratis code, but they didn't know the cost
of bad gratis code.
One faction of OSI began with X.25 as its Layer 3, and
added the minimal-function Transport Classes 0-3. That was the PTT/IBM side, of
course. The peer-networking side demanded and got Connectionless Network
Protocol (CLNP) and Transport Class 4. This was inspired by TCP/IP, of course,
but was far superior! It took into account what was learned in TCP/IP. TP4's
syntax is far cleaner than TCP's. This part of OSI was held back by problems
elsewhere in the stack.
In the early 1990s, there was an effort at IETF to
build a new IP to replace IPv4, whose address space was looking tight even then
(mostly due to inefficient classful assignment). IAB voted for "TUBA", which
was really a profile of OSI CLNP -- it had already been implemented on the major
routers and hosts of the day, because OSI was always looming on the imagined
horizon. Then Vint changed his mind and went to the hacked combo of Steve's IP
and Paul's IP (SIP and PIP) that became IPv6. Over a dozen years later and it's
still not widely implemented, thank heavens, because it's not worth the effort.
It was anti-OSI prejudice, a big "NIH", that killed TUBA, which could have been
in place before the Internet boom got going.
The upper layers were a
disaster. Layers 3 and 4 were basically done by the early 1980s. Session and
Presentation got in the way, and some of the Applications being written were too
complex. Eventually the upper layer folks figured out Application Service
Elements, putting optional functions into Layer 7. But TCP/IP had completely
captured the mainstream market before Layer 5 and 6 were fixed.
But they
were fixed. OSI ended up with a little known "fast byte" protocol for both
Session and Presentation. Four bits in the first packet signaled that Layers 5
and 6 would have a packet header length of 0. It was a clever political
compromise -- the old-model purists, folks who never worried about
implementation and didn't want to admit to a mistake, could still claim that the
layers were present, but realists noticed that they were, uh, inactive.
OSI
was sort of a good idea, but having one project do what were essentially two
parallel stacks (I like that two towers and a garaga model) was not a good one.
Two really parallel stacks, unencumbered by each other, would have done
better.
The sad thing is that TCP/IP is not all that good -- it is just
better than what the OSI folks could turn out (see my previous comment about
Rosecode, which was intended to illustrate, not implement, the protocols). Both
are obsolete. But because OSI is seen as being the evil spawn of PTTs and TCP
as the good guys, nobody wants to look at the fundamentals and do it up right.
It's not as if we haven't learned anything since 1973. But TCP/IP has gotten
"to Moses on Sinai" status, which it doesn't deserve. It has served us
reasonably well. It really has become the open standard for interconnection.
But it is being stretched well beyond its efficient scope, and has more patches
on it than, well, Apache (y'all know how that got its name?) ever
could.
BTW, not to be too picky, but I'm always suspicious of an article
that talks about ISO without knowing that the group's English name is
"International Organizaton for Standardization". "ISO" is the TLA precisely
because it is not the true acronym in any of its main languages, and because of
the prefix "iso-". [ Reply to This | # ]
|
| |
| Authored by: Anonymous on Sunday, February 20 2005 @ 11:52 PM EST |
"The perfect is the enemy of the good."
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Monday, February 21 2005 @ 02:45 AM EST |
| The members of ISO are holding meetings to divide up the power to control
the
global Internet.
It matters not that they already failed once. Because they
hold the reins of
power, they fully intend to wield it. Over the next few
years, there will be
power grabs. The reward is that great.
History has a
way of repeating itself. Think monopolies. [ Reply to This | # ]
|
| |
| Authored by: elderlycynic on Monday, February 21 2005 @ 05:23 AM EST |
It really peeves me when people who win out politically use
their position to deprecate the losers. While 90% of that
is fact, the remaining 10% is biassed in a way that makes it
offensive to the people who worked on OSI's predecessors.
I was not one, and so am speaking up for them.
Firstly, OSI's predecessor in design (the UK JNT Coloured
Books protocols) was working and in-use at about the same
time as TCP/IP. There were advantages and disadvantages to
both, but the Coloured Books were at least as reliable, and
used between a MUCH wider range of systems. TCP/IP never
caught up with that, largely because most systems nowadays
are Unix-like.
Yes, the primary reason that OSI flopped was that its
bureaucracy got out of hand, and it got nowhere fast, but
there was another. The USA government pushed its "Star
Wars" funding into computer science research (Project
Athena and all that), and TCP/IP got a MASSIVE financial
boost that its competition didn't. This is why we have the
ghastly X Windowing System, rather than the technically
superior NeWS.
Lastly, TCP/IP is NOWHERE near as reliable as its proponents
make out. Without putting it down, those of us with
experience of both systems remember the vastly superior
diagnostics (and, to some extent, reliability) of the
Coloured Books with nostalgia. I could go into this area
at length, but it is too deeply technical for Groklaw.
[ Reply to This | # ]
|
| |
| Authored by: chris_bloke on Monday, February 21 2005 @ 05:37 AM EST |
Caveat: I'm going to use the terms X.25 and X.29
interchangeably, because at the time that's what we did
and I can't
for the life of me summon up the enthusiasm to
figure out which was
which, 15 years or so on.. :-)
I started at university
in 1987 in Aberystwyth, and
within a few months fell in with a bunch
of folks writing
a bulletin board on the Honeywell L66 that was
there. This
evolved into AberMUD
As part of all this I rapidly became aquainted
with the
Joint
Academic
Network (JANET), an X.25 based network run
by the JNT<
/A>
that connected all the UK universities. Over this ran the
various colo
ured
book protocols, such as Greybook
for email and the blue book file transfer
protocol. We
had dumb terminals with serial connections to
X.29 PAD
and could then place calls to any of the JANET machines,
if you knew the X.25 address. This was relatively easy as
the
name to X.25 address mappings were published in big
text files which
you could get to by making an X.25 call
to uk.ac.aber.info
(which our PADs had been
programmed to know) and look things up there.
Of course, there were other things to do,
like the Bullet
BBS at UCL (with whom we had a very friendly rivalry)
and play MIST:
CALL 000049600001
LOGIN 2653,2653
Ahem, I seem to have gotten distracted.. Anyway, as time
went on and I
struggled on trying to balance my physics
degree and my newfound interest
in computing, I got
heavily involved with the Computer
Unit at Aberystwyth, building a
n
X.400 mail system (PP) as part of the JNT's (doomed)
X.400
pilot. This was against the background of the decision
in 1985 to move
JANET to OSI from X.25 when the OSI
standards became
stable. This
raised the spectre of losing the UK
domain for the ISO country-code compliant
GB, something
that a lot of people thought was crazy. At least with a
transition to Internet addressing all the rest of the
world needed to do was
reverse our address (though that
caused problems for many sites like
uk.ac.aber.cs).
As
others
have said X.400 was tbe best excuse for A4 business
cards
ever.
But then in a matter of a months the JIPS (Janet IP
Service)
project was born, tunnelling TCP/IP over X.25 via
routers and suddenly things
like DNS, Telnet and SMTP mail
became possible!
A few years
after that and it had all changed, first of
all from TCP/IP over X.25 to X.25
over TCP/IP (XOT) and
then (in 1997) the last remnants of the X.25
service was
terminated.
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Monday, February 21 2005 @ 08:44 AM EST |
"And there truly is no connection between this honor and spam."
You're missing the point. Mr. Cerf's actions since inventing TCP/IP make him
unworthy of the honor being bestowed. If someone does bad things, it's
generally considered inappropriate to bestow honors on them for other things
they've done. No matter what good he's done in the past, it doesn't excuse his
behavior of today.[ Reply to This | # ]
|
| |
| Authored by: justjeff on Monday, February 21 2005 @ 11:53 AM EST |
Only in 1994 was the [OSI] FIPS finally
rescinded.
Even in 1994 OSI wasn't quite dead yet.
At the time
I worked for a US government agency who was not exactly ahead of the technology
curve. In 1994 and 1995 the agency finally admitted to itself that networking
computers had value. Even as NBS/NIST was finally concluding that OSI should be
withdrawn in favor of TCP/IP, my agency put out RFPs for networking. My agency
was always ready to believe that really big companies knew what they were
talking about, believed the really big telephone companies (the only commercial
players still hoping for some X.25 business).
Even in 1994 and 1995, my
agency spent millions on OSI networks and X.25 equipment. It never really
worked. It never really gatewayed very well to the real internet. Eventually
even my agency pulled the plug and replaced the whole mess with TCP/IP
networks.
The funniest part is, the OSI lesson was never really learned.
Yes, my agency uses TCP/IP, but not the way it should be used. My agency's
telecomm group, who live and breath whatever the big telephone companies tell
them, installed TCP/IP routers, but then disabled their routing abilities,
replacing them with excruciating lists of predefined fixed routes.
The
telephone guys, to this day, cannot see the flaw in their TCP/IP network
design. All of the "routers" are nothing more than multiplexors. In the eye of
a telephone guy, I guess this is the way it should be.
The consequences
are obvious. The wildly complex fixed routing tables are often lost or
corrupted or erroneously edited by various staff. Network outages are
commonplace. Network repair and reconfiguration can take hours and
hours.
There was once a magazine article called "Real Programmers don't eat
Quiche." One of its themes was that FORTRAN programmers could write FORTRAN
programs in any language. My agency's theme would be, OSI network engineers can
implement OSI network phiosophies in any network.
- jeff -
[ Reply to This | # ]
|
| |
| Authored by: Anonymous on Monday, February 21 2005 @ 01:17 PM EST |
"...I think this is relevant and important because, though it took over 15
years, technology won. The "good guys," led by Cerf, Kahn, Walden, and
others defeated the governments, PTTs, and giant corporations by using
well-engineered, open protocols -- stuff that worked and was robust."
He was describing TCP/IP. I've heard the same said of the Linux. So now we
know where Linus' place in history will be in a few years.[ Reply to This | # ]
|
| |
| Authored by: Eagle on Monday, February 21 2005 @ 02:52 PM EST |
Some are reportedly thinking of writing a letter of formal complaint
(...) because of perceived issues with MCI, Cerf's employer, and spam,
protesting the award to Mr. Cerf. I hope they don't embarrass themselves with
such an ungrateful, misguided, and unkind gesture. Spam is awful, but it is not,
I must say, the only awful thing people can do to one another. Internet thuggery
is plenty awful too. Mr. Cerf should be allowed to enjoy this wonderful honor,
which is well-deserved, in peace. (...) And there truly is no connection between
this honor and spam. It's just not nice to ruin this moment, and I hope they
will reflect and desist.
Just to clarify what these issues are: Spamhaus claims that MCI is
allegedly profiting from spam, however this is Vint
Cerf's recommendation for responsible use of the Internet (published as early as
1994-08-14), and that's what he is
doing at MCI. He is also quoted as saying
this:Spamming is the scourge of electronic-mail and newsgroups on
the Internet. It can seriously interfere with the operation of public services,
to say nothing of the effect it may have on any individual's e-mail mail system.
... Spammers are, in effect, taking resources away from users and service
suppliers without compensation and without authorization.
Which
begs the question how much private anti-spam "enforcement" can be
demanded from an ISP (facing potential liability if it terminates a
contract with immediate effect but without readily available and watertight
proof for the alleged involvement of its customer with spam gangs) in a country that, as
Spamhaus concedes, has made the mistake of legalizing much of the spam.[ Reply to This | # ]
|
|
|
|
|
