On Mon, Apr 8, 2013 at 5:36 PM, Clem Cole <clemc at ccc.com> wrote:
On Mon, Apr 8, 2013 at 5:05 PM, Johnny Billquist <bqt at softjar.se> wrote:
The flag day for TCP/IP was 1 Jan 1983, so I wouldn't expect you were
running much TCP/IP before that point. (Yes, I know experiments and
development was going on, but the number of implementations were few, still
had issues, and was very much work still in progress
Johnny it was TCP/IP. Remember, I'm one of the implementors of the original
IP/TCP for the VMS (along with Stan Smith) in >>1979<<. I was also 3Com
first customer at the same time (another but related story). Most people
do not realize the first product 3Com sold was >>software<< - UNET a TCP/IP
implementation for UNIX/V7 (PDP11 and Vax) - we took deliver on Dec >>32<<
1979 because 3Com had a funding thing with their VCs that they would ship
before the end of 1979.
I would hardly call IP/TCP a work in progress. Yes, it was young, but it was
well defined. Most of the major sites had switched and the US Gov had a
spent a bunch to make sure it was implemented. We had it running on a
number of interesting and different systems at the time. If I had the time
and can actually read the tapes, at one time I >>had<< the bits on 9-track
for many of them in my basement (I still have the tapes - but who knows).
FYI: the original IP/TCP for 4.1BSD was not written at Berkeley, it was
written at BBN and used the MIT Chaos-Net hacks to slide in the 4.1BSD
kernel (by Rob Gerawitzs & Rob Walsh). Remember, BBN had the contract from
ARPA to develop the different IP/TCP implementations. In fact, the mbuf
code that Rob G created was because he needed a memory handler that was OS
kernel independent, so it could be stuffed into a number of a different
kernels. Eric Cooper was the grad student that put the "portable BBN
IP/TCP" into 4.1 at UCB to replace the BerkNet and Eric Schmidt (yes the
Google one) made the mailer talk to it. Berkeley had a contract to support
the base UNIX kernel for ARPA. So as part of that, wnj would create
"sockets" for 4.1A (as a response to the Accent/Mach "port" concept) and
then re-stuff the BBN code into his socket layer. Then he, Sam, et al start
to hack it. Van would take it up the hill to LBL and start to hack
further. Eventually 4.2BSD would be released as we know it as part of the
UCB ARPA contract and most sites picked up the code from that release not
the BBN release.
DEC all of these release along the way and Fred Canter, Armando Stettner,
and the whole "TIG" (telephone industries group) in Merrimack were doing
their thing for AT&T, the Universities and any UNIX licensee that wanted it.
TIG would begat the Ultrix team.
Not trying to come down on you, but "I was there" and very much "mixed up"
in it all.
As for when MOP was released for the UNIX flavors, I really can not
remember. It was all around the same time, but as I said, those bits in my
brain are stale and I was not part of any LAT/MOP etc (directly or
indirectly) so their is no real reason for me to remember some of the
specifics.
Clem
Hello!
Clem, I (not speaking for Johnny, but speaking for myself and four
cats. Also responsible for the perennial problem surrounding Dave.)
would never heap stuff on you. I actually applaud your efforts. As for
the stuff that still fastens the Internet together, it is indeed a BBN
solution. In fact in Cliff Stoll's book he mentions that bit of trivia
and also what to expect from them.
However he does complain as to how the routing of the Internet works.......
-----
Gregg C Levine gregg.drwho8 at gmail.com
"This signature fought the Time Wars, time and again."
On 2013-04-08 23:26, Paul_Koning at Dell.com wrote:
On Apr 8, 2013, at 5:09 PM, Johnny Billquist wrote:
...
By the way, MOP, from an ethernet point of view, is neither LAT nor DECnet. MOP is its own protocols on ethernet. 0x6001 and 0x6002. But I believe they were defined as being a part of the DECnet suite anyway.
The MOP spec is one of the DECnet architecture specs. And in fact it relies on the DNA datalink layer and is controlled via the DNA management layer. Other than that, it's indeed separate -- doesn't use routing or NSP.
Paul, what do you mean by "relies on the DNA datalink layer"?
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
On Mon, Apr 8, 2013 at 5:05 PM, Johnny Billquist <bqt at softjar.se> wrote:
The flag day for TCP/IP was 1 Jan 1983, so I wouldn't expect you were running much TCP/IP before that point. (Yes, I know experiments and development was going on, but the number of implementations were few, still had issues, and was very much work still in progress
Johnny it was TCP/IP. Remember, I'm one of the implementors of the original IP/TCP for the VMS (along with Stan Smith) in >>1979<<. I was also 3Com first customer at the same time (another but related story). Most people do not realize the first product 3Com sold was >>software<< - UNET a TCP/IP implementation for UNIX/V7 (PDP11 and Vax) - we took deliver on Dec >>32<< 1979 because 3Com had a funding thing with their VCs that they would ship before the end of 1979.
I would hardly call IP/TCP a work in progress. Yes, it was young, but it was well defined. Most of the major sites had switched and the US Gov had a spent a bunch to make sure it was implemented. We had it running on a number of interesting and different systems at the time. If I had the time and can actually read the tapes, at one time I >>had<< the bits on 9-track for many of them in my basement (I still have the tapes - but who knows).
FYI: the original IP/TCP for 4.1BSD was not written at Berkeley, it was written at BBN and used the MIT Chaos-Net hacks to slide in the 4.1BSD kernel (by Rob Gerawitzs & Rob Walsh). Remember, BBN had the contract from ARPA to develop the different IP/TCP implementations. In fact, the mbuf code that Rob G created was because he needed a memory handler that was OS kernel independent, so it could be stuffed into a number of a different kernels. Eric Cooper was the grad student that put the "portable BBN IP/TCP" into 4.1 at UCB to replace the BerkNet and Eric Schmidt (yes the Google one) made the mailer talk to it. Berkeley had a contract to support the base UNIX kernel for ARPA. So as part of that, wnj would create "sockets" for 4.1A (as a response to the Accent/Mach "port" concept) and then re-stuff the BBN code into his socket layer. Then he, Sam, et al start to hack it. Van would take it up the hill to LBL and start to hack further. Eventually 4.2BSD would be released as we know it as part of the UCB ARPA contract and most sites picked up the code from that release not the BBN release.
DEC all of these release along the way and Fred Canter, Armando Stettner, and the whole "TIG" (telephone industries group) in Merrimack were doing their thing for AT&T, the Universities and any UNIX licensee that wanted it. TIG would begat the Ultrix team.
Not trying to come down on you, but "I was there" and very much "mixed up" in it all.
As for when MOP was released for the UNIX flavors, I really can not remember. It was all around the same time, but as I said, those bits in my brain are stale and I was not part of any LAT/MOP etc (directly or indirectly) so their is no real reason for me to remember some of the specifics.
Clem
On Apr 8, 2013, at 5:09 PM, Johnny Billquist wrote:
...
By the way, MOP, from an ethernet point of view, is neither LAT nor DECnet. MOP is its own protocols on ethernet. 0x6001 and 0x6002. But I believe they were defined as being a part of the DECnet suite anyway.
The MOP spec is one of the DECnet architecture specs. And in fact it relies on the DNA datalink layer and is controlled via the DNA management layer. Other than that, it's indeed separate -- doesn't use routing or NSP.
paul
On 2013-04-08 23:05, Johnny Billquist wrote:
On 2013-04-08 18:13, Clem Cole wrote:
I'm scratching very stale memory bits here, but Ultrix for the Vax/750
circa 1982/83 might have had a MOP server that did not need a full
DECnet install. We did not run DEC-Net at one of my jobs because we
had TCP/IP which did everything DECNet could do and worked across
vendors. But I thought had a couple of LAT devices to support some
dial-up modems and a funky DEC printer that after it booted, spoke TCP
fine, but used MOP to boot. I was not involved with configuration or
maintaining any of it, so none of the details ever stuck and lend
together in my mind. But I would think you look at Ultrix around that
time you might find something.
Ultrix definitely got a mop server at some point. If it was around in
1982/83 is something I can't comment much on, however...
My memories of LAT were that it was it's own protocol in the ethernet
sense (i.e. had it's own packet type IIRC I want to say 6003) which was
different from what DECnet used. It was a very low overhead protocol,
very MIT Chaos-Net like, and much more efficient for terminals and other
low speed devices than TCP.
Correct. LAT is a protocol on ethernet. 0x6004 in fact. 0x6003 is
actually DECnet.
LAT is rather efficient, but it don't have any routing capability.
By the way, MOP, from an ethernet point of view, is neither LAT nor DECnet. MOP is its own protocols on ethernet. 0x6001 and 0x6002. But I believe they were defined as being a part of the DECnet suite anyway.
Johnny
Johnny
Clem
On Sun, Apr 7, 2013 at 8:07 AM, Bob Armstrong <bob at jfcl.com
<mailto:bob at jfcl.com>> wrote:
> LAT is _NOT_ part of DECnet. LAT can be run without DECnet
installed
> or running. IIRC, LAT was licensed with VMS.
Yeah, but most (actually, "all" I think) of the DECservers
required MOP to
download them. How did you do that without DECnet?
Or was it just an unwritten catch-22 that you had to have a
DECnet node
somewhere to boot up your terminal server, even if it wasn't the
node you
actually wanted to connect to ?
Bob
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
On 2013-04-08 18:13, Clem Cole wrote:
I'm scratching very stale memory bits here, but Ultrix for the Vax/750
circa 1982/83 might have had a MOP server that did not need a full
DECnet install. We did not run DEC-Net at one of my jobs because we
had TCP/IP which did everything DECNet could do and worked across
vendors. But I thought had a couple of LAT devices to support some
dial-up modems and a funky DEC printer that after it booted, spoke TCP
fine, but used MOP to boot. I was not involved with configuration or
maintaining any of it, so none of the details ever stuck and lend
together in my mind. But I would think you look at Ultrix around that
time you might find something.
Ultrix definitely got a mop server at some point. If it was around in 1982/83 is something I can't comment much on, however...
The flag day for TCP/IP was 1 Jan 1983, so I wouldn't expect you were running much TCP/IP before that point. (Yes, I know experiments and development was going on, but the number of implementations were few, still had issues, and was very much work still in progress.)
NCP maybe?
I know we had a 3Com terminal server that was a piece of work, but we
ran the UUCP link of the Vax for a long time and the modems I thought
were on a LAT. I just don't remember.
My memories of LAT were that it was it's own protocol in the ethernet
sense (i.e. had it's own packet type IIRC I want to say 6003) which was
different from what DECnet used. It was a very low overhead protocol,
very MIT Chaos-Net like, and much more efficient for terminals and other
low speed devices than TCP.
Correct. LAT is a protocol on ethernet. 0x6004 in fact. 0x6003 is actually DECnet.
LAT is rather efficient, but it don't have any routing capability.
Johnny
Clem
On Sun, Apr 7, 2013 at 8:07 AM, Bob Armstrong <bob at jfcl.com
<mailto:bob at jfcl.com>> wrote:
> LAT is _NOT_ part of DECnet. LAT can be run without DECnet installed
> or running. IIRC, LAT was licensed with VMS.
Yeah, but most (actually, "all" I think) of the DECservers
required MOP to
download them. How did you do that without DECnet?
Or was it just an unwritten catch-22 that you had to have a
DECnet node
somewhere to boot up your terminal server, even if it wasn't the
node you
actually wanted to connect to ?
Bob
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
I'm scratching very stale memory bits here, but Ultrix for the Vax/750 circa 1982/83 might have had a MOP server that did not need a full DECnet install. We did not run DEC-Net at one of my jobs because we had TCP/IP which did everything DECNet could do and worked across vendors. But I thought had a couple of LAT devices to support some dial-up modems and a funky DEC printer that after it booted, spoke TCP fine, but used MOP to boot. I was not involved with configuration or maintaining any of it, so none of the details ever stuck and lend together in my mind. But I would think you look at Ultrix around that time you might find something.
I know we had a 3Com terminal server that was a piece of work, but we ran the UUCP link of the Vax for a long time and the modems I thought were on a LAT. I just don't remember.
My memories of LAT were that it was it's own protocol in the ethernet sense (i.e. had it's own packet type IIRC I want to say 6003) which was different from what DECnet used. It was a very low overhead protocol, very MIT Chaos-Net like, and much more efficient for terminals and other low speed devices than TCP.
Clem
On Sun, Apr 7, 2013 at 8:07 AM, Bob Armstrong <bob at jfcl.com> wrote:
> LAT is _NOT_ part of DECnet. LAT can be run without DECnet installed
> or running. IIRC, LAT was licensed with VMS.
Yeah, but most (actually, "all" I think) of the DECservers required MOP to
download them. How did you do that without DECnet?
Or was it just an unwritten catch-22 that you had to have a DECnet node
somewhere to boot up your terminal server, even if it wasn't the node you
actually wanted to connect to ?
Bob
On Apr 7, 2013, at 8:07 AM, Bob Armstrong wrote:
LAT is _NOT_ part of DECnet. LAT can be run without DECnet installed
or running. IIRC, LAT was licensed with VMS.
Yeah, but most (actually, "all" I think) of the DECservers required MOP to
download them. How did you do that without DECnet?
Or was it just an unwritten catch-22 that you had to have a DECnet node
somewhere to boot up your terminal server, even if it wasn't the node you
actually wanted to connect to ?
Bob
Correct.
paul
On Apr 6, 2013, at 7:40 PM, Brian Schenkenberger, VAXman- wrote:
"Cory Smelosky" <b4 at gewt.net> writes:
Hmmm. There is no LATCP.EXE, LAT$STARTUP.COM, or LTLOAD.COM it would
appear. Was it an add-on kit for VMS 3.5 or did I miss something rather
important? ;)
Though I lived it, recalling this history is not one of my strong points.
LAT showed up in VMS circa 1984/1985 time frame from my recollection. I
think you will need a newer version of VMS if you want LAT. Look around
for V4.7 (pre-SMP/V5.0) or later VMS. They should have LAT and DECnet IV.
I think LAT is older than that, but it certainly is part of the Ethernet era, so that would be early 1980s. Not 1979, which is the date on that V1.50 release. No Ethernet, so no LAT.
paul
On Fri, 5 Apr 2013, Paul_Koning at Dell.com wrote:
On Apr 4, 2013, at 6:44 PM, Clem Cole wrote:
...
But that's for terminals up to 9600 baud. For networking, you'd use a DMC-11 unless your OS supported the cheaper devices and money was that tight -- that one goes back to about 1976 and delivers up to 1 Mb/s depending on model (up to 56 kB/s long haul, given suitable modems).
Hmm - are the DMC and DMR that old? I remember fighting the firmware in the them to allow high speed serial networking. They had a dedicated microprocessor on them (8080A or 8085 IIRC - but it may have been something custom). They were expensive, which why Berk-NET used 9600 baud serial lines, until we got 3COM & Interlan Ethernet cards at Berkeley in 1983.
My 1976 Peripherals Handbook has a description of the DMC-11 in it. And I also remember it was used in Typeset-11, which had a custom network implementation specific to it that included Phase III style routing back in 1978. (No relationship to DECnet at all either in architecture or implementation.)
A DMC-11 is essentially a KMC-11 with programming fixed in ROM, rather than dowloadable in RAM, plus a line card. The KMC-11 processor is a custom engine, its instruction set looks somewhat like microcode. No connection to any Intel chips, that couldn't possibly have come within a mile of the performance requirements. Come to think of it, the first use of a 808x series chip in DEC products I can think of is the head servo control processor in the RA80. There may have been 8031s in some other spots, I no longer remember where I saw those.
...
But I do remember our college main timesharing system, in 1973, a PDP-11/20 with 28 kW of memory, RSTS V4A, and 16 terminals on 16 separate KL11 or DL11 interfaces. Oh yes, and a mean time between crashes of about 1 day.
Are you sure it was a 11/20, not an 11/40? I did not think RSTS could run without the MMU. With 16 DL/KL11's even with an 11/40 the interrupt rate had to been wretched.
Positive. It was RSTS V4A-12, which did not use an MMU and required only 28 kW of memory (24 kW for a minimal install). RSTS started requiring an MMU in version 5, the first version that was called RSTS/E (for "Extended" as in extended memory).
And yes, a box full of single line serial cards. Most of them ran at 110 Baud driving ASR33s; one or two were talking to TI Silent 700 terminals (300 baud printing on thermal paper), and one was feeding a Beehive editing terminal, don't remember what speed, almost certainly no higher than 1200 Baud because it was about 1000 feet from the computer. All this on 20 mA current loop connections, no RS232 as far as I remember. (The Beehive was for the London Stage project, an amazingly complex project to digitize and index a large body of historical reference books, back when OCR didn't really exist yet. There's a neat book about it, "Travels in Computerland" by the project director prof. Ben Schneider.)
Also an RK05 for system disk, an RF11 swap disk, and some DECtapes for additional file storage in case you wanted to save more stuff than could fit on the "large disk" (i.e., the RK05).
Because of the reliability issues, we had a long battle with DEC to get it fixed. The actual root cause may have been interference from the nearby campus FM broadcast transmitter (3 kW around 90 MHz). But whatever it was, we never got a real fix for that; eventually DEC threw in the towel and delivered a "replacement part" -- an 11/45 with a pile of new peripherals and a RSTS/E kit. :-) That did the job.
paul
Our 11/20 (also running RSTS V4A-12) ran for years with only 1 LA36 tied to it. There were 2 DL11's in the chassis (not hooked to anything) and 2 old ASR33's in the basement, but it was obvious these had not been hooked up in years. It was strange to inherit a machine that had a multitasking, multiuser operating system on it, with only 1 terminal (running 300 baud). The math professor that had been in charge of it was leaving and needed someone to take command of this fine old machine. We couldn't wait until he FINALLY gave us the system 1,2 account password which was: SECRET
Yes, it had never been changed, and he had never run a sysgen. And unlike
Paul's experience, I don't think this machine ever crashed. When we did finally appropriate the funds to purchase 2 VT55 terminals and hooked the DL11's to them, it was like watching paint dry, waiting for the screen to update at 300 baud. My Elmer found some information on installing larger capacitors and a different pot that would allow us to run 2400 baud. This was quite a chore as we were supposed to be able to dial it all in with an oscilloscope (that never did get it to work). Eventually, I wrote a small display loop that was directed at that DL11 and we manually turned the pot while it was running until the gibberish turned into something readable. The large caps on the DL11's held the cards in the BA11 at strange angles and we always thought they would short together, but somehow, it all ran.
Our original disk drives were the Diablo RK03's which were not very reliable (we couldn't wait to upgrade to 3 RK05's). That extra RK05 with an additional 2.5 meg made all the difference in the world. ;-)
Brett
