Hello tenser;

tenser wrote to N1uro <=-

On 23 Aug 2021 at 09:03p, N1uro pondered and said...

I just checked my FreeBSD 13.0-RELEASE-p3 system, and the only mention
I find of Phil Karn in the networking code is comment referring to his retransmit algorithm, which first appeared in a 1987 paper at ACM
SIGCOMM while he was at Bellcore. To my knowledge, he never worked at NEC.

I'm going back to the earlier versions... 3,4,5.

The string KA9Q doesn't appear in the kernel source or header files at all. There is a brief reference to "Karn's rule" in clearing errors in the SCTP implementation, but it's not obvious that that refers to Phil Karn.

Of course. KA9Q is his amateur callsign and may not be referenced, however
his name is. Even a "rule" is an influence.

Karn is, of course, well known in technical circles and I know for a
fact he knows many of the people who worked on the BSD TCP stack, but
it doesn't look like there is any NOS code in the BSD stack.

You misunderstood me. NOS code may not be in BSD however his developmental rules are... as you pointed out above. NOS is for DOS... not for unix or a variant there of. Having worked with K2MF on his MFNOS, I'm pretty familiar with the NOS code as a lot of it was still Phil's work and/or theory. He's
even mentioned in the PBS documentary "birth of the internet".

... Terminal Illness - Getting sick at the airport
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

Since a couple asked...

Once at NEC Corporation two engineers, one who was a ham, were having a friendly
discussion about how to link a PC to their corporate VAX system. The non-ham actually ended up challenging the ham to come up with a way to communicate to it. Not one to back down from a good challenge, the ham came up with a means for the PC to run a program on 5.25" floppy disk using less than 250K DOS ram to do so with.

After figuring out how to connect as a client, the ham (KA9Q) also figured out that he could run tcp/ip server services within the same executable without giving up much more ram. Such services included NNTP and SMTP which another
ham friend of his developed. Also FTP and later on the amateur packet radio protocol ax.25.

This system written by KA9Q/Phil Karn is called NOS and was placed in the public domain. If you read deeply in various UNIX/BSD and Windows code you'll see KA9Q NOS all over both IP stacks. Since then other hams have continued development on the code such as Brian Lantz with TNOS, Barry Siegfried with MFNOS (which I've worked on with him) and still under development is JNOS2
now maintained by Maiko Langlaar.

Hams have their own IP network in which they administer and handle themselves, some, like myself, have migrated to being dual stacked both on public internet AND over amateur radio... but for the PC it all started at NEC Corporation
by a dare to talk IP to a VAX.

... Tumour - One plus one more
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

WOW! That's pretty rad, and I never knew HAM's had their own IP range they administer. HAH.. time to hit the google machine. So much I never knew! =)

|15Z|07ylone

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: pLANET cARAVAN BBS (21:3/150)

Hey Zylone;

Zylone wrote to N1uro <=-

WOW! That's pretty rad, and I never knew HAM's had their own IP range
they administer. HAH.. time to hit the google machine. So much I never knew! =)

Hams are the true mothers of invention. Recently they sold off the last /10
of their /8 subnet and have been using that money now to pay certain individuals and give grants and scholarships however they're headed in more
of an IPv6 direction since there's more IPs and it's NOT supposed to route under ax.25 even though I've engineered a way for it to do such. For now I personally coordinate for 1/5th of the U.S.A. which has me in control over
10 /16 subnets. Seems funny that my /48 IPv6 block has many many more IPs
than their original /8 had.

... Virginity is experience not yet fulfilled
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

On 19 Aug 2021 at 05:10p, N1uro pondered and said...

This system written by KA9Q/Phil Karn is called NOS and was placed in the public domain. If you read deeply in various UNIX/BSD and Windows code you'll see KA9Q NOS all over both IP stacks.

That is absolutely not true. BSD sockets were developed
entirely independently of KA9Q's stack, and I'm quite sure
Phil would corroborate that if asked.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

Hey tenser;

tenser wrote to N1uro <=-

That is absolutely not true. BSD sockets were developed
entirely independently of KA9Q's stack, and I'm quite sure
Phil would corroborate that if asked.

This is dependant on which variant of BSD. Reading through some of the header files, on some versions of FreeBSD it specifies credits towards the KA9Q NOS. Phil's NOS has been public domain for a while now.

... Labour Pain - Getting hurt at work
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

On 23 Aug 2021 at 09:03p, N1uro pondered and said...

That is absolutely not true. BSD sockets were developed
entirely independently of KA9Q's stack, and I'm quite sure
Phil would corroborate that if asked.

This is dependant on which variant of BSD. Reading through some of the header files, on some versions of FreeBSD it specifies credits towards
the KA9Q NOS. Phil's NOS has been public domain for a while now.

What versions?

I just checked my FreeBSD 13.0-RELEASE-p3 system, and the only mention
I find of Phil Karn in the networking code is comment referring to his retransmit algorithm, which first appeared in a 1987 paper at ACM SIGCOMM
while he was at Bellcore. To my knowledge, he never worked at NEC.

The string KA9Q doesn't appear in the kernel source or header files at
all. There is a brief reference to "Karn's rule" in clearing errors in the SCTP implementation, but it's not obvious that that refers to Phil Karn.

Karn is, of course, well known in technical circles and I know for a fact
he knows many of the people who worked on the BSD TCP stack, but it doesn't look like there is any NOS code in the BSD stack.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

On 25 Aug 2021 at 03:22a, tenser pondered and said...

all. There is a brief reference to "Karn's rule" in clearing errors in the SCTP implementation, but it's not obvious that that refers to Phil Karn.

Ah, I just checked this; that comment doesn't indeed refer
to Karn's retransmission algorithm. There is a paper on
making SCTP robust with respect to spurious transmissions
that cites the 1987 ACM SIGCOMM paper by Karn and Partridge,
and that seems to be the origin of that comment.

Still, this is all most definitely independent of the KA9Q
TCP/IP stack.

I can find no evidence that any KA9Q code ever made it into
a Unix-based TCP/IP stack, let alone is pervasive in the BSD
implementation. Karn's work was good, but it was independent.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

Hello tenser;

tenser wrote to N1uro <=-

On 23 Aug 2021 at 09:03p, N1uro pondered and said...

I just checked my FreeBSD 13.0-RELEASE-p3 system, and the only mention
I find of Phil Karn in the networking code is comment referring to his retransmit algorithm, which first appeared in a 1987 paper at ACM
SIGCOMM while he was at Bellcore. To my knowledge, he never worked at NEC.

I'm going back to the earlier versions... 3,4,5.

The string KA9Q doesn't appear in the kernel source or header files at all. There is a brief reference to "Karn's rule" in clearing errors in the SCTP implementation, but it's not obvious that that refers to Phil Karn.

Of course. KA9Q is his amateur callsign and may not be referenced, however
his name is. Even a "rule" is an influence.

Karn is, of course, well known in technical circles and I know for a
fact he knows many of the people who worked on the BSD TCP stack, but
it doesn't look like there is any NOS code in the BSD stack.

You misunderstood me. NOS code may not be in BSD however his developmental rules are... as you pointed out above. NOS is for DOS... not for unix or a variant there of. Having worked with K2MF on his MFNOS, I'm pretty familiar with the NOS code as a lot of it was still Phil's work and/or theory. He's
even mentioned in the PBS documentary "birth of the internet".

... Terminal Illness - Getting sick at the airport
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

In regards to the linux kernel:
linux/tcp.h:375: * Slow start and congestion control (see also Nagle,
and Karn & Partridge)
linux/rslib.h:9: * RS code lifted from reed solomon library written by Phil Karn
linux/rslib.h:10: * Copyright 2002 Phil Karn, KA9Q

Just a couple of examples of Karn's influence on protocol stacks on other OS platforms.

... Tablet - A small table
--- MultiMail/Linux v0.52
* Origin: Carnage - risen from the dead now on SBBS (21:4/107)

On 24 Aug 2021 at 01:01p, N1uro pondered and said...

I'm going back to the earlier versions... 3,4,5.

It's unclear to me what you are trying to say. Your
original message said that if you look, you see,
"KA9Q NOS all over" the BSD and Windows TCP/IP stacks
and something about hams at NEC (Phil Karn never worked
for NEC, AFAIK) and a VAX.

What I said is that you do not. You see references
to his 1987 paper on TCP retransmission, coauthored
with Partridge.

In particular, you did not refer to Phil Karn's papers
or research work, you referred specifically to his
ham-centric KA9Q software package. As I said, I see
no evidence that KA9Q NOS had any influence on the BSD
TCP/IP stack. Obviously Karn's other work did have
influence (and continues to).

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

I can find no evidence that any KA9Q code ever made it into
a Unix-based TCP/IP stack, let alone is pervasive in the BSD implementation. Karn's work was good, but it was independent.

Your not wrong... or right. :) The problem is your going back to the beginnings of Internet which means no internet to distribute source bases.
It was done (like early works of GNU was on mag tapes) but generally that wasn't how we code shared back in the day.
IN GENERAL we code shared by explaining how something worked and the more complicated stuff we would print it on a printer and send it to the other person. I have no clue what Karn did or did not do. BUT he very well might have sent a print of some or all of his work and sent it to Berkeley team who would than have to re-type it into their code base. If it's their code base they very well might not have credited Karn. A lot of things were just given back than and wasn't thought about that 30 years from now someone would care. We all imagined hover boards and jetpacks and everything we worked on would
be pointless by than.
I assure you if they did use Karn's code no one was trying to discredit him, just no one thought about it at all.
But ya... in those old days the only way to be a contributor of a code base
was to basically be physically there in the building.
Pre-BSD sockets really is pre internet as we know it. So your talking about
a time where print outs were the most common way to share code.

--- Mystic BBS v1.12 A46 2020/08/26 (Raspberry Pi/32)
* Origin: Darkest Hour BBS (dhbbs.org) (21:3/145)

On 24 Aug 2021 at 01:30p, BadOPCode pondered and said...

I can find no evidence that any KA9Q code ever made it into
a Unix-based TCP/IP stack, let alone is pervasive in the BSD implementation. Karn's work was good, but it was independent.

Your not wrong... or right. :) The problem is your going back to the beginnings of Internet which means no internet to distribute source
bases. It was done (like early works of GNU was on mag tapes) but generally that wasn't how we code shared back in the day.
IN GENERAL we code shared by explaining how something worked and the more complicated stuff we would print it on a printer and send it to the other person. I have no clue what Karn did or did not do. BUT he very well might have sent a print of some or all of his work and sent it to
Berkeley team who would than have to re-type it into their code base.
If it's their code base they very well might not have credited Karn. A lot of things were just given back than and wasn't thought about that 30 years from now someone would care. We all imagined hover boards and jetpacks and everything we worked on would be pointless by than.
I assure you if they did use Karn's code no one was trying to discredit him, just no one thought about it at all.
But ya... in those old days the only way to be a contributor of a code base was to basically be physically there in the building.
Pre-BSD sockets really is pre internet as we know it. So your talking about a time where print outs were the most common way to share code.

Nope, sorry; with respect to how Unix development was done in
the 1970s and 1980s this is completely inaccurate.

There really is no KA9Q NOS code in the BSD TCP stack.
Independent of code, I don't believe there was any significant
influence on the BSD stack, either.

Karn and Partridge published a paper at ACM SIGCOMM in 1987 on
calculating TCP round trip times; some techniques from that
paper are implemented in most modern TCP/IP implementations
(though they aren't strictly speaking necessary, and have
lead to some problems on really high-latency networks), but
that is qualitatively different: Karn was a researcher working
for Bellcore at the time, and while he was (and is!) an avid
amateur radio operator, and he did write KA9Q, he was also
someone who was quite literally paid to do this kind of
research, independently of his work on KA9Q.

As far as TCP/IP in Unix goes, the stack actually originated
at BBN, who did a namei-based hack using the "standard" Unix open/close/read/write for 4BSD under contract for DARPA: The
DoD wanted to standardize on a common software system for its
contractors, and Unix on the VAX won (the other major competition
was VMS on VAX; the PDP-10, which had been the major machine
type on the ARPANET and on which some early TCP/IP work had
been done, was nearing EOL at the time and was out). Anyway,
the BBN stack made it back to Berkeley, where it was bug-fixed
and enhanced by Bill Joy et al and put back into the BSD
distribution; this lead to something of a kerfuffle with BBN
who had continued development of their stack under the assumption
they were going to provide the final software; UCB wanted to
ship what had basically become sockets. There was a shootout
between the stacks, and the BBN code paniced on the test machine
while the Berkeley code did not, so BSD sockets became the
"official" DoD reference implementation of TCP/IP. This was
all done before 1983, as 4.2BSD included TCP/IP. This is all
documented in Salus's book on Unix history and the 'net, and in
Lyon and Hafner's books on Internet history ("Where Wizards Stay
Up Late").

According to Phil Karn's web page at http://ka9q.net/code/ka9qnos/,
KA9Q NOS began life in 1985, several years after TCP/IP shipped
in 4.2BSD, and it started on CP/M. Incidentally, according to that
page, MIT had done a PC IP implementation earlier.

I'm sorry, but the assertion that KA9Q NOS had much influence on
ether the Berkeley Unix TCP/IP stack or Winsock is just not based
in fact.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

I'm sorry, but the assertion that KA9Q NOS had much influence on
ether the Berkeley Unix TCP/IP stack or Winsock is just not based
in fact.

I'm sorry but your wrong with your assertion. I was never arguing with you
EVER about Phil's involvement with BSD TCP stack. Just that it's very
possible he could have shared info or snippets of code as it was a popular thing to do than.

As far as old ways of code sharing. I was a kid but I remember it. It was mostly done mostly with print outs or rarely punch cards being mailed. Who else remembers drawing and making cool things out of punch cards? By mid-80's things had gone to mostly exchanging by mailing floppies. BUT large projects were usually still not distributed as entire code bases and would still be
just key pieces parts.
The original BSD/MIT open source licensing was if you use this code leave me out of it. Everyone was more concerned being fingered for damaging a corp or government mainframes than worried about getting kudos. But everyone shared a lot up until a famous jackass from Seattle hit the scene and would claim he invented electrons if he could squeeze another dollar out of you.
But BSD has recieved a lot of code share from a lot of surprising players (yes even the jackass.) They just were rarely complete works like they are today.
I really think Stallman was the first with GNU to really push the distributing of complete works. IMHO. BSD and MIT ... hard to explain but it was all just a weird experiment to these guys. Like no really cared much of the overall state things were in outside their systems. Not a fact, just the feeling I got back than. But anyways... the number of developers working on data telecomunications was really small and everyone knew everyone and a lot of people loved sharing with each other information.
So it's hard to say who the unsung contributors on the information and
concept level of things for software made in the 80's. Tons of code sharing both snippets and complete works exploded with the advent of the BBS. Double that, when PC's got internet and not just terminals to mainframes on the internet.
But as far as Phil Karn working at Berkeley writing their TCP stack as an uncredited contributor... very much doubt that. And back than that is what it would have taken. To physically be there. But was he ever in a room with the Berkeley guys after a few beers which is ultimately what I was talking
about... and I have no clue.
BUT if you were to talk to the jackass from Seattle he invented the BSD
socket stack right after he invented the operating system, GUI's and oxygen.
;)

--- Mystic BBS v1.12 A46 2020/08/26 (Raspberry Pi/32)
* Origin: Darkest Hour BBS (dhbbs.org) (21:3/145)

On 24 Aug 2021 at 08:28p, BadOPCode pondered and said...

I'm sorry, but the assertion that KA9Q NOS had much influence on ether the Berkeley Unix TCP/IP stack or Winsock is just not based
in fact.

I'm sorry but your wrong with your assertion. I was never arguing with
you EVER about Phil's involvement with BSD TCP stack. Just that it's
very possible he could have shared info or snippets of code as it was a popular thing to do than.

What, precisely, is the assertion that you claim I made that is wrong?

How could a program Phil Karn wrote for CP/M in 1985 have influenced a
VAX Unix code base shipped in 1983? Recall that the genesis of this conversation were claims that a) the first TCP/IP stack on a microcomputer
was written by a Karn (not true; the first was PC/IP, by Karn's own
admission), b) that Karn worked at NEC at the time (not true; he worked
either at Bell Labs in Illinois or at Bellcore by then), c) that "KA9Q
NOS" is "all over" the TCP/IP stacks shipped with BSD Unix and with
Windows (there is no evidence to support that assertion). Sure, the BSD
TCP/IP implementation has been updated over the years, but bear in mind
it was the reference implementation and by that time, people working on
such things had access to the Internet, had had access to the ARPANET
before that, had access to UUCP and would have been attending multiple
relevant conferences per year.

As far as old ways of code sharing. I was a kid but I remember it. It
was mostly done mostly with print outs or rarely punch cards being
mailed.

Sorry, no. Work on Unix, the Internet, etc, was mostly being done
on networks by that point. In the limit, people were mailing around
magtapes (or hand-carrying them to conferences, such as Ken taking
a copy of Unix to Berkeley in 1975). But things like this were being
passed over UUCP and the ARPANET very early on.

Certainly, no one working on those systems were mailing around decks of
punched cards.

This is all well documented. I recommend Salus's book, "Twentyfive
Years of Unix": he speaks with most of the major players.

Who else remembers drawing and making cool things out of punch
cards? By mid-80's things had gone to mostly exchanging by mailing floppies. BUT large projects were usually still not distributed as
entire code bases and would still be just key pieces parts.

What you are describing was, perhaps, true in the microcomputer/hobbyist
world, but not on minicomputers and not in the research community.

The original BSD/MIT open source licensing was if you use this code
leave me out of it. Everyone was more concerned being fingered for damaging a corp or government mainframes than worried about getting
kudos. But everyone shared a lot up until a famous jackass from Seattle hit the scene and would claim he invented electrons if he could squeeze another dollar out of you.

Not quite. In the very early days, computers were all about the
hardware, and software was mostly an after thought; this is why the
very early IBM mainframe OS's are all publicly available: IBM gave
them away because they didn't see a value add. IBM started charging
for software when Amdahl got an injunction that basically said that
IBM had to support his software on their machines.

Unix is a more interesting story; the Bell system was a regulated
monopoly, and in exchange for their monopoly status they had to give
away their research artifacts for a "reasonable fee." This meant
that anyone from a university or research center could basically ask
for a copy for the cost of media and shipping, etc, though AT&T
retained the rights. This led to a rapid proliferation of Unix
on relatively inexpensive minicomputers in places like university
academic departments, which could now afford a computer of their
own (a mini was like 10x less than a mainframe).

By the time of the 7th Edition release (1979), AT&T was starting to
tighten up its licensing terms, hence why Lyons didn't do another
edition of his famous "papers" which described 6th Edition in detail.
The VAX was coming on the scene by then and it's proprietary operating
system, VMS, was proprietary, which was the norm at that point.

But BSD has recieved a lot of code share from
a lot of surprising players (yes even the jackass.)

Microsoft was an early Unix licensee, but I'm not aware of much, if
any, MSFT code that made it into BSD, at least not until post 4.3BSD,
when POSIX was really ramping up. MSFT was much more closely involved
with System V; SVR4 merged System V, Xenix, SunOS, and parts of BSD.
There are a smattering of SVR4 descendants still in use today (AIX,
HP/UX, and Solaris). Of those, only the Solaris descendants are
open source.

They just were rarely complete works like they are today.

Incorrect. By 3BSD, UCB was distributing a complete Unix
distribution to AT&T licensees. The usual pattern was to buy
a VAX without VMS, get a 32/V license from AT&T, and then
write to Berkeley for copy of BSD.

I really think Stallman was the first with GNU to really push the distributing of complete works.

No. GNU never distributed a complete system. Despite
Stallman's claims that it should be referred to as "GNU/Linux"
they never had a kernel until Linux came along and de facto
adopted the GNU utilities and environment. Before that, it
was common to install GNU software on commercial Unix (they
fixed a lot of bugs that existed in the code bases that made
up System V and BSD distributions).

IMHO. BSD and MIT ... hard to explain but it was all just a weird experiment to these guys. Like no really cared much of the overall
state things were in outside their systems. Not a fact, just the feeling
I got back than.

Sorry, that's not true; by the mid-1980s, Unix on minicomputers
and the emerging workstation market were a big economic factor
in the computer industry. Again, refer to Salus's book.

But anyways... the number of developers working on data
telecomunications was really small and everyone knew everyone and a lot of people loved sharing with each other information. So it's hard to say who the unsung contributors on the information and concept level of things for software made in the 80's.

I think you're mostly referring to microcomputers and hobbyist
development, which was very different.

Tons of code sharing both
snippets and complete works exploded with the advent of the BBS.

This is where a lot of folks commenting from the hobbyist perspective
get bogged down: there was a vibrant research community that predated
the BBS by a few decades that was doing serious work in this area,
but unless you were involved in that community, you probably didn't
know much about it. Sharing there often happened by papers published
in journals, code and patches shared by e.g. UUCP or the ARPANET,
and face-to-face interactions at conferences (never underestimate
the power of the hallway track).

Double that, when PC's got internet and not just terminals to mainframes on the internet.

Personal workstations started emerging in a serious way towards the
mid 1980s, and many of those were directly connected to the Internet,
well in advance of the mass introduction of PCs.

But as far as Phil Karn working at Berkeley writing their TCP
stack as an uncredited contributor... very much doubt that. And back than that is what it would have taken. To physically be there. But was

No, sorry, that's just wrong. As I've said, there was communication
via other means. As a research, Karn would have had access to much
greater facilities for such things than the average hobbyist.

he ever in a room with the Berkeley guys after a few beers which is ultimately what I was talking about... and I have no clue.

He certainly presented at conferences they attended. They definitely
added support for the KP87 retransmission algorithm to the Berkeley
TCP/IP stack. But, as I said before, that's qualitatively different
than KA9Q NOS being "all over" the BSD and Windows code.

BUT if you were to talk to the jackass from Seattle he invented the BSD socket stack right after he invented the operating system, GUI's and oxygen. ;)

No, Microsoft had to obey the terms of the BSD license; they credit
the UC regency for Winsock. Xerox invented the GUI as we know it from
both Apple and Microsoft.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
* Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)