Close encounters of the ITS kind

larsbrinkhoff commented 4 years ago

Taking notes about ITS "almost" getting out of MIT.

Except for the KS10 version, which did get out and about.

larsbrinkhoff commented 4 years ago

Richard Schroeppel wrote:

I was back visiting the lab around 1980, and asked if I could take a copy of ITS and the software we used (Teco, etc.) back to the Foonly group where I worked in Culver City. They gave me a tape, after some discussion. I wound up not using it.

larsbrinkhoff commented 4 years ago

Date: Tue 13 Mar 84 23:58:47-PST
From: Ken Harrenstien <KLH@SRI-NIC>
Subject: ITS on a Foonly?
To: ai-kl%oz@MIT-MC
cc: klh@SRI-NIC

In light of the recent discussion about buying a $99K 2020, I decided
not enough was known about the Foonly alternative and sent some
questions to Tymshare.  I am distributing the responses I received,
with some editing of extraneous material; I have also included my
original questions in brackets where it seems necessary for clarity.
--KLH

----------
Date: 29-Feb-84 01:41 PST
From: Robert N. Lieberman  <RLL.TYM@OFFICE-2>
Subject: Re: Foonly machine & code availability
To: Ken Harrenstien <KLH@SRI-NIC>

Ken, I guess I the right person to ask most of those questions. let me try ...

[KLH: Are F2s or F3s still available?  What about the newer machines?]

The F2 (or F3 as we call it) is no longer made.  We have  8 but they are all
used at TYM. Some clients some internal users.

There are two models of interest relating to F4.

System 26 is based on the F4 CPU and runs TENEX. It has ucode that
supports the KA instruction set and emulates the KA TENEX map (pager).
We have several of these, one is at SRI and one at CECOM in New Jersey
(both of these are owned by customer).  This model is no longer being
made although one might be obtainable in the future.

System 26KL is the latest model. It is also based on the F4 CPU but is
modified to support the full extended addressing (actually 32 bits
hardware-wise but that is not used since TOPS-20 does not use all
those bits either). The 26KL has totally different ucode and emulates
the KL instruction set including ALL the extended business instruction
set. it emulates the KL style map.  It runs the TYMCOM-20 operating
System which is user identical to the TOPS-20 system. The 26Kl is
spec-ed at 2.5 times the KA at the machine instruction level.  This
seems to be proved out but we are undergoing a big performance push
right now and hope to have formal data by spring showing this to be at
true.

System 26KL is MOST definitely available for leasing or buying.  We
are gearing up for production of these beasts.  There is a strong
backing internally to use these machines for internal use in addition
to external sales/leases. This provides a built in demand.

the cost of a nicely configured system (1 meg memory, a few  disks (660 MB),
etc.) is about $300,000.  You could probably push on that for universities but
not sure how much.

Yes, we are willing to provide access to ucode but this has yet to be
made a firm committment from upper management. The front end (if used
in the KL sense) does NOT exists.  There is a Console computer that is
used for booting and diagnostic work but has no relationship with I/O
functions of the mainframe.

We have TYMNET but no MILNET yet since we do NOT have the TCP/IP
sources for TOPS-20 (can we get that from someone?? we do have a
source license for TOPS-20 so I think we are entitled to it).

We are anxiously seeking out programmers for both the monitor level
work and for ucoding. I have two opening right now.  We have hired a
few enginerring types to help with manufacturing and for enhancements.
I personally would love to have cooperative ventures with
universities.

Let me know if there is more interest from any quarters.

Robert

-------------------
Date:  2-Mar-84 23:47 PST
From: Robert N. Lieberman  <RLL.TYM@OFFICE-2>
Subject: Re: Foonly machine & code availability
To: Ken Harrenstien <KLH@SRI-NIC>

[KLH: I asked for more details on the speed differences...]

The 26KL has NO cache and although it is on the list of possible items
to do, we currently have NO plans to implement it.  You first have to
understand why a cache on the 2060 is worth it. The memory cycle time
is something like 800 ns so a cache can have a very dramatic effect.
The 26KL has a memory cycle time of 350 to 425 depending on how we set
some parameters.  Right now we can run with no problems at 375 ns.
Thus, we feel the cache will NOT provide a significant enhancement of
performance as compared to some other ideas.

The 26KL instruction speed (running the KL set) varies from lower than
the KL (I won't say figures) to higher than the KL. On the average
(what ever that means) it seems to come out to about 60% of the KL.
Remember this is an instruction set speed comparison.  I think the
2060 wins bigger in the I/O due to multiple controllers. Currently two
is the max for the 26KL.

We have a very nice list of wonderful things we can do with the 26kl
to make it much better (VERY likely better than the 2060) but one
OVERRIDING factor MUST happen.  WE MUST SEE IF THE MARKETPLACE WILL
ACCEPT THE 26KL.  (basically as is).  If not it does not make sense
for us to develop it.  We are rounding it out now and cleaning it up.
It really is there and works. Several are at TYM now.  We are very
anxious to see the marketplace show interest.  It did at the show but
that seems to have died off.  Very disappointing. (of course you may
find fault with the lack of heavy advertising on our part, regretably
that is a fact but not sure if that would have changed the situation).
Whatever the marketplace says, TYM is very committed to produce these
machines for internal use if for not other reason. However, future
effort will key off of marketplace acceptance.

The complete instruction set is there, full addressing is there. A
wonderful implementation of the KL map is there.  Power, size and cost
seem to heavily favor the 26KL.  With TYM maint., we can provide the
whole thing.

[KLH: I asked whether the price could be cut by leaving off the
suede leather trim, racing stripes, etc.]

Now for costs. certainly a barebones CPU (with 1/2 meg memory) can go
for much less than the 300k figure. I will leave that up to the
marketing types. Yes, i heard the 250k rumor but when you add
everything else, it is easily 600k or more. So we are still 50% of
their [KLH: DEC-2060] cost.

Hope this helps.

Robert

---------------------
Date:  4-Mar-84 23:36 PST
From: Robert N. Lieberman  <RLL.TYM@OFFICE-2>
Subject: Re: Foonly machine & code availability
To: Ken Harrenstien <KLH@SRI-NIC>

I would contact Cal Peterson for marketing info. I don't have his
number here at home but will send it to you later.

[KLH: I asked about the status of unbundling TOPS-20 -- not relevant
to ITS, but interesting.]

As for software on the 26KL, it is a matter of legalese on what
TYMCOM-20 is.  Certainly TOPS-20 cannot run on the 26KL and TYMCOM-20
cannot run on the DEC equipment.  We have very different architecture
which means parts of the monitor must be VERy different. however, from
the user view, we are as identical as one can be. So what is it??  We
claim, with some legal opinion, that TYMCOM-20 is NOT TOPS-20 and
hence no licensing is needed by us or our customers.  We certain do
realize this might not completely dispell the fear of some legal
action by DEC so we are trying hard to talk to DEC to clarify all
this. TO date they are VERY hardnose.  I think everyone agrees they
are trying to delay things with the hope we just go away or die.

My point of view says this is very short sighted for DEC, bad for
their customers, and extremely sad for all those fans of the DEC 36
bit line.  I sincerely wish to join forces with DEC and have mutual
software develoment for the 36 bit line. DEC can do somethings and TYM
can do others.  Clearly the DEC 36 bit machines are now in the end of
the line phase.  If DEC wants to keep its base and hopes to move them
onto the DEC 32 bit line, they need to do more.  Joining with TYM will
be a GIANT step forward. DEC has NO hope to convert those 36 bit sites
to VAXen in the next 2 years.  Why not hold onto them by letting them
use a TYM 26KL.  Eventually (if we are to believe DEC) the VAX will be
bigger and better and more convenient for a 36 bit to VAX move. As I
view it, we are really saving DEC!!!!

Robert
-----------------------

KLH: Well, it doesn't look as if a Foonly will be vastly cheaper than
a DEC machine, but it does seem to have more potential than a 2020.
The microcode differences may or may not make it harder to modify the
paging; again, I can't judge this.  If anyone wanted to pursue the
cooperative angle, it may be possible to get a much better deal,
although it's not clear whether Tymshare would expect improvements to
TYMCOM-20 rather than to the machine, interfacing, or microcode.  If
for some reason anyone wanted to bundle together an ITS for export
outside MIT, a 26KL would be an attractive direction to go in.  I just
wish they were faster about getting the installed base in place.