Talk:IMP interface

From Computer History Wiki
Revision as of 18:14, 13 November 2024 by Jnc (talk | contribs) (→‎DR11C: I think the SRI interface was originally designed for use with the DR11-C)
Jump to: navigation, search

DR11C

Re DR11C. I peeked at the driver in 2.11BSD and I think it said Unibus. But it was only a brief glance. Larsbrinkhoff (talk) 10:11, 15 March 2018 (CET)

Hey, this mentions UNIBUS:

Larsbrinkhoff (talk) 10:19, 15 March 2018 (CET)

Huh? That page talks about the IMP11-A, the DEC CSS thing.
Anyway, for the SRI thing, it was possibly both, actually. What it was was an SRI board that took the bit-stream from the IMP, doing the host-IMP hardware protocol ('there's your bit', etc), and converted it to words, which it shipped over a parallel interface to a standard DEC DRV11 card. I'm pretty sure the QBUS DRV11 and UNIBUS DR11-C had the same parallel port spec, so you could probably have plugged the SRI card into a DR11-C instead of a DRV11. Since the DR11-C/DRV11 are programmed I/O, they wouldn't have had the performance of the others, which were DMA, which is probably why UNIBUS machines tended to go with the DEC/ACC interfaces. Jnc (talk) 15:55, 15 March 2018 (CET)
I have a vague memory that the early SRI interface manual talks about the DR11-C; i.e. that interface was originally designed for use with the DR11-C. The DRV11 was just a later plug-compatible substitution. Jnc (talk) 18:14, 13 November 2024 (CET)
Sorry, wrong link. This is better:
Yes, I see now DR11-C is the name of a parallel interface. The BSD drivers use the Unibus device to talk to the IMP interface.
Here's a manual for the IMP interface:
Larsbrinkhoff (talk) 08:36, 16 March 2018 (CET)
Right. And here is the 2.11 driver:
Looking at the driver, I'm not sure I understand how it works; it looks like it might loop in the interrupt handler, reading the entire packet? Eh, not important.
Somewhere I have MOS operating system drivers for it.
Also my memory was a bit off - it was byte at a time, not word at a time. Jnc (talk) 15:14, 16 March 2018 (CET)

Dynamic Modeling

MIT-DMS wasn't a typo. Although the situation is confused with many different names over the years (DMCG is another prominent one), I believe the official ARPANET name was MIT-DMS.

Here's a line from a 1980 MIT hosts list: 

HOST MIT-DMS,           1/6,SERVER,ITS,PDP10,[DM,MITDM,MIT-DM,DMS]

So MIT-DM and plain DM were acceptable aliases. Other lists, e.g. https://github.com/ttkzw/hosts.txt, only says MIT-DMS with no aliases. Larsbrinkhoff (talk) 05:32, 21 October 2021 (CET)

We always called it plain 'DM', and a lot of software called it that, too - e.g. MLDEV. (IIRC, there was a directory which held binary for loadable devices, so when you referenced 'XXX:AAA; BBB CCC' it went and looked there for the correct file for XXX - the filename format and directory escape me at the moment - but you could look at a dump and see if there's a 'DMS' entry there, as well as 'DM'; they were actual files, so the dump would have captured them.) So that's why I changed it (my only goal is maximum accuracy), but I don't have any major commitment to using 'DM'; if you feel that 'DMS' is more accurate, feel free to change it back. Jnc (talk) 13:22, 21 October 2021 (CEST)
The file names for loadable devices are "DEVICE;JOBDEV xxx" where 'xxx' is the device name.
But while looking for that, in SYSDOC;ITS RECENT I found this:
Date: Fri, 28 Mar 86 09:02:47 EST
From: Alan Bawden
All kind of worms are crawling out of the woodwork because of various programs that -know- that all ITS machines are named "AI", "MC", "ML", or "DM".
which matches my memory (above). Jnc (talk) 16:41, 21 October 2021 (CEST)
Yes without a doubt the software on ITS called it DM. It was my intent to use the ARPANET name, as used in the official host lists. But reviewing the article, it's not so important in this context.

More interfaces

Two more interfaces: one for ANTS, and one made at ISI. Stephen Casner wrote: My first work at ISI as a grad student was to help with the hardware debugging of the newly built ISI IMP Interface, a variant of the ANTS Imp Interface from Illinois. At the end of EPOS deployment the aforementioned 11/44 had interfaces to the ARPAnet and the Wideband Satellite Network that were implmented in UMC-Z80 add-in devices developed by Lincoln Lab.

This also hints at the story that ANTS was made so Illinois researchers could access Illiac IV. Larsbrinkhoff (talk) 07:11, 28 October 2021 (CEST)

Yeah, I saw that the ANTS guys did a (UNIBUS) 1822 interface. I didn't try to find any information on it. I wonder how (if at all) it relates to the IMP11-A interface from DEC? That must not have been around when the ANTS interface was done (otherwise why do to the work). I'm trying to remember who used the IMP11-A - ISTR that BBN had a number of them. (And if I get really ambitious, I'll get my scanner machine working again; I should scan my IMP11-A manual, and put it online.) The UIUC UNIX NCP code and the BBN UNIX V6 should show which 1822 interfaces those used.
Also, in the MOS-related investigations, it turned out that SRI co-developed with ACC a DMA 1822 interface. Whether QBUS or UNIBUS I don't recall - maybe first one, and then the other. (The earlier Stanford/SRI interface that used the DR11/DRV11 was interrupt per character, so not good for high-volume applications like the Port Expander - the later Port Expander product used the ACC interface). I have a couple of ACC 1822 manuals here, I should look at them and write up something about those interfaces. (And make sure they are online. Bitsavers has little ACC stuff; the UNIBUS 1822 manual they have seems to have come from me :-). I'll have to look and see if the MIT Gateway used the Stanford or ACC interface; I don't recall. Jnc (talk) 16:08, 28 October 2021 (CEST)
OK, the TIU manual (which I found in machine-readable form, so I won't have to scan it - it's not available on the Interwebs) says "Associated Computer Consultants, under contract to SRI, designed and fabricated an 1822 interface based on a multichannel microprogrammable DMA controller". And the Port Expander manual describes it as a 2+ board set - a "Micro-11 DMA controller board" and 1-5 "1822 .. interface boards". So that's not the LH-DH/11 Local/Distant Host Controller‎, which must have been done earlier. (The LH-DH/11 is all SSI TTL chips.)
The Stanford gateway ('Golden' - which I maintained for some years; it used the MIT CGW code) used the DMA interface, and the main MIT gateway used the DRV11 interface; the latter machine doesn't appear to have ever switched (but see below). There was also a 'Test' ARPANET gateway at MIT, which sometimes used one, and sometimes the other. Odd! I have no idea why - maybe we stuck with the DRV11-based one on the main MIT gateway because we had spares for that, and we had the DMA one on the Test gateway to make sure that config worked, for Golden.
A potential fly in all this ointment is that I moved out to Proteon, and did a lot of PDP-11 MOS improvements there; e.g. I used the memory management unit on the -11/23 to move all the packet buffers into high memory, allowing i) lots more buffers, and ii) more room for code (which all had to live in the bottom 56K). I'm pretty sure that code was used on the MIT machines (the 3Mbit Ethernet card, which only MIT had, could only do DMA to the bottom 64K, so I had a lot of hair with a separate special 'low buffer pool' in the bottom 64K, which the 3Mbit Ethernet card used; packets that came in on another card and were headed out the 3Mbit Ethernet card had to be copied down - but maybe that was all done for Golden). I have printed listings for a lot of that code, and at least one tape, but I have not gotten it read yet, and I'm not sure either will have per-machine config info in it. Jnc (talk) 14:49, 30 October 2021 (CEST)
OK, I have downstairs the 'LH-DH/11 Maintenance Manual' (but a different one from the one in BitSavers which has my initials on it :-), the preliminary 'XQ/1822 Maintenance Manual', and 'MDMA Multichannel DMA Controller for LSI-11 Maintenance Manual'. Maybe more upstairs? Jnc (talk) 18:29, 29 October 2021 (CEST)
I found ELF source code and it has support for the ANTS/ISI interface (they're almost the same). It's a simple DMA device, see https://github.com/larsbrinkhoff/simh/issues/11
netbot.p11 says THE MCCONNEL IMP INTERFACE USED AT SRI-AI. ORIGINAL VERSION WAS DESIGNED TO USE THE UNIV. OF ILL. ANTS INTERFACE. From the source, it should be possible to write up the programming spec for all these. Jnc (talk) 21:04, 29 October 2021 (CEST)
KS10 ITS uses the/an ACC 1822 Unibus interface, and it uses DMA. Maybe the same? Larsbrinkhoff (talk) 18:27, 29 October 2021 (CEST)
The same as which one? The ANTS or ISI or McConnell/SRI? Or the SRI/ACC DMA interface I mentioned above? (Saying 'SRI interface' isn't specific enough, there were at least two (the DR11 one, and the SRI/ACC DMA one).
I'm not sure if the SRI/ACC DMA interface was the LH-DH/11, which was a UNIBUS device; the SRI Port Expander product (which probably, but not definitively, used the SRI/ACC DMA interface) was a QBUS machine. I'll have to look and see if I have the MOS driver for the SRI/ACC DMA interface, I'm pretty sure I do.
Of course, they all might be program-compatible, in which the drivers won't tell us anything definitive. Jnc (talk) 21:04, 29 October 2021 (CEST)
The QBUS DMA card (the 'MLH-DH/LSI11', the manual calls it) appears to be pretty much program compatible with the LH-DH/11.
I have this sneaky feeling that the MOS DMA driver was written for the LH-DH/11, as I suspect a lot of this MOS stuff was running on UNIBUS machines before the QBUS gear all showed up. Jnc (talk) 14:49, 30 October 2021 (CEST)
Probably not. The LH-DH/11 has a bit in the transmit CSR which isn't there in the MLH-DH/LSI11 - "WC0". The MOS 1822 DMA driver defines all the bits - except that one. Also, a lot of the bit names changed between the LH-DH/11 and the MLH-DH/LSI11 (even though they do the sae things!) - and the MOS 1822 DMA driver uses the new names. Jnc (talk) 23:51, 30 October 2021 (CEST)
So I looked at that [ANTS/ISI interface stuff]; it's pretty clearly neither an IMP11-A, nor an LH-DH/11 clone. Jnc (talk) 16:11, 4 November 2021 (CET)
Thanks! As for KS10 ITS, I meant the ACC LH-DH Unibus card. That's how ITS and KLH10 refer to it, anyway. Larsbrinkhoff (talk) 16:41, 4 November 2021 (CEST)
Interesting; the LH-DH/11 manual clearly indicates it's in a 10-1/2" box - which I don't remember one on either the AI or MC KS's. Unless there was a later version on a single board, which could mount in the BA11 mounting box. I'll have to ask Alan (and maybe JTW remembers too). Jnc (talk) 18:30, 4 November 2021 (CET)

Duplicative?

I guess this is kind of duplicative, now that 1822 interface‎ exists. Maybe not, though; that page is focused on the 1822 interface spec itself (for use in places like Packet Radio Network), whereas this page is more about the various host/bus interfaces that produce 1822 interface capability. I guess this distinction/separation isn't as clear as it could/should be, currently; I'm busy at the moment, but I'll look at these in the next day or so. Jnc (talk) 16:38, 8 November 2021 (CET)

I also noted this, but was thinking "IMP interface" is more a view from the host side. Now that I read the page contents here, I think it could just as well be a section on the 1822 interface page. Larsbrinkhoff (talk) 17:37, 8 November 2021 (CET)
I suppose there wouldn't be any loss of generality in covering the 1822 interface spec and the host interfaces that support it on one page. I have to go ponder all this, though. Jnc (talk) 20:03, 8 November 2021 (CET)
I guess it's almost a coin-flip whether to have two pages (divided between the interface spec, and host interfaces that implement it) or two. I don't have the energy to do a merge, in part because there are lots of pages that have been adjusted to point to one or the other; it's easier to just leave things the way they are. I have looked at this page to make sure it's cleaned up to follow the 'division of content' scheme. Jnc (talk) 15:29, 13 November 2021 (CET)

Alto interface

There's enough info on the Alto 1822 interface in the Alto 1822 archive to do a good page on it, which I will do. Jnc (talk) 15:29, 13 November 2021 (CET)

Retrieved from "https://gunkies.org/w/index.php?title=Talk:IMP_interface&oldid=35493"