USA Visit March 1982

Bob Hopgood

• 1. SIGGRAPH Papers Committee
• 2. Three Rivers
• 3. Carnegie-Mellon

1. SIGGRAPH

The first two days were spent at the SIGGRAPH Papers Committee in New Hampshire. This was a long meeting - we worked from 8.30 am to 2.30 am one day! With me being jet lagged as well, I effectively got to bed at 7.30 am!

SIGGRAPH expect about 7,000 delegates to the conference in Boston. It is still growing at an almost absurd rate. Counting attendees at the Exhibition, it should go well over 14,000.

There were about 160 papers from which they wished to accept about 30 - consequently, quite a lot of good papers were turned out and a lot of anguishing over which should stay in.

The tutorial programme is enormous and is now attracting as many delegates as the conference itself. There are 6 different tutorials this year on Business Graphics alone. They would like to have a GKS Tutorial in 1983 and would like to use the people giving it at EUROGRAPHICS 82. Peter Tanner has the names and will contact individuals.

Major news items associated with the Single User Systems work (called Professional Workstations in USA now) were:

1. General view that PERQ was dead now that the SUN terminal has appeared. Having said this, there is a general lack of awareness of where PERQ has got to and the need for somebody to build SUN terminals in quantity. It is the same kind of euphoria that PERQ had two years ago. Major complaint/worry is that PERQ is far too expensive (SUN terminal will cost $7,000), no software, inability to track microprocessor advances quickly etc.
2. Jim Clarke's geometry chip is now working and will be available on the SUN terminal. Gives 3D rotation, clipping and proximity checking.
3. We can order a SUN terminal now if we wish to - Jim Clarke is contact - company is SUN Terminals Inc, or something like that. I have two unpublished papers on the geometry chip and the SUN rasterop. As I should not have them, I would prefer people to borrow rather than copy.
4. Bob Sproull is working hard on anti-aliasing techniques for grey level displays. Has one or two algorithms based on storing grey level lines at certain angles and interpolating between them rather than regenerating to improve speed of production.

2. THREE RIVERS

This day was a bit messed up by Ed Fredkin arriving from Boston unexpectedly and calling a Board meeting. No idea what the panic was about - nobody seemed to know.

I visited the new production facility and advanced development laboratory on the RIDC Industrial Park on the outskirts of Pittsburgh and spent time talking to Don Seltzer, John Strait, Miles Borell, Jim Marshall (new head of development) and Jim Gay.

The new production facility is still in the middle of being set up. The factory used to be owned by Simplicity - the pattern maker. They are still in the process of building offices, laying cables, installing phones etc.

It is a very big area - did not get the square footage - but more than they need at present. The ten PERQs under test looked a bit lost in the middle. All of the equipment that was down at Gross Street seems to have moved out there -the board placement, flow solder, board tester etc.

Main points of interest from the visit were:

The PERQ shells with ICL printed on them are made in Pittsburgh so we were deluding ourselves when we thought they indicated UK manufacture. All 10 systems under final test when I was there were 50 cycle UK systems with ICL on them. They are identical with the USA versions apart from the power supply (Boschert-Sunnyvale) and some changes that have to be made to satisfy European safety standards. The person who showed me round said they did batches of one or the other as there was sufficient difference to be careful. Some of the systems were 0.5 Mbyte and some 1 Mbyte. 3RCC have shipped ICL a number of systems with 0.5 Mbyte and 1 Mbyte memories which ICL said they required for SERC. You can tell the amount of memory that the system has by looking at the end of the memory board. The 0.5 Mbyte and 1 Mbyte have it written on. The 0.5 Mbyte have every other row of RAMs in place.

The VMI tubes are no longer modified by 3RCC but are changed to 3RCC specification by VMI themselves. 3RCC have no second supplier at the moment but are confident that they can get all the tubes they need from VMI.

The aim is to get MTBF down to 1 per year!

The person who showed me round mentioned that they had delivered 3 ethernet boards to IBM Cambridge Scientific Centre and they had 3 systems running on an ether with a Canon Laser Printer attached. They will have ethernet demo software running at SIGGRAPH where you draw on the tablet of one system and it comes up on the screen of the other.

At the Advanced Design Laboratory, I did not get a lot of information. Brian Rosen had already left for UK. Most of this information came from Don Seltzer:

1. Multi processor version of D6 is working. The next major release will be a system with D6 functionality based on the SPICE kernel. No dates.
2. They have a CANON laser printer with ethernet software that allows rudimentary FTP to the PERQ that is acting as a print server. I have an example of the output. Two programs they have, output a file or dump the screen.
3. They were not too happy with ICL's decision to go with the ECMA ethernet superset. Some of the fields used for special purposes in this are rather eccentric (type field?).
4. They had the CMU DP program up and running under D6. They were using it in earnest on board design. Boards were designed using it and simulated by the TEGUS 5 package which they have running on a VAX. Has cut board design times from 1 month to 1 week. Point was made that most designs in USA are done on A4 sheets and the PERQ screen was quite comfortable for engineers to work with.

   Don Seltzer was working on a program to display waveforms - effectively a logic analyser.

   John Strait was working on a new drawing program which had a more natural hierarchical structure with the user able to define his own profile. Would be aimed at being used with one hand on tablet and one only on keyboard. Calling it EGAD - Editor for Graphical Data.

5. They have no plans for a window manager - would use CANVAS. Did not seem to know that HCR were supposed to be doing one.
6. CMU are doing a LISP implementation again!
7. They are working on a program for measuring performance. Checks where a program spends all its time.
8. D6 includes a full set of trigonometric functions - 32 bit floating point.
9. Advanced Development had turned over the TRUST LIBRARY to marketing/sales.
10. Nobody working on ISO PASCAL.

I had lunch with Jim Marshall and found him quite impressive. Apparently 3RCC had tried to split off the immediate developments from Brian Rosen's group in the past. They had found a good person to take the job but Brian insisted that he report to Rosen. The person eventually decided that it would not work and refused the offer.

Since then the company has persuaded Brian that they would not get anyone unless the person was independent of the Advanced Development set up. Jim Marshall was recruited for this post.

Jim Marshall comes from DEC where he was in charge of VAX developments. DEC had two projects going in parallel - one to do a GATEARRAY 2-board system and the other a l-board VLSI system. DEC decided to cut the 2-board project and made the whole team of about 12 people redundant.

3RCC had decided to set up in Boston as it was easier to recruit staff there. Jim Marshall had set up home in Pittsburgh while keeping his house in New Hampshire. They had envisaged the facility in Boston would not be set up for a year or so. They managed, however, to recruit the whole of the DEC team who all have relevant experience and know how to work with each other. Consequently, the move of development to Boston is likely to happen some time this year.

This group will be responsible for all developments going out in the field in the short term while Rosen's group will look at the full 32-bit system. It looks as though low cost PERQ (TEACUP) will be done by Rosen's group.

Most of the new developments are dependent on the new I/O board etc which will arrive with the KRISTMAS system (something for everybody). This will have the new 8" MICROPOLIS disc, new I/O board, noise down to 55 DB etc. This will be done by the Boston group. Will have a single large fan.

The MICROPOLIS disc is currently 45 Mbyte unformatted. Due to go to 90 Mbytes and there will be room for two in the chassis.

In the meantime, Jim Marshall is keen to get some projects underway which could come out before KRISTMAS which will not be until the end of the year. He is seriously looking at a landscape display (they have a reasonable monitor) and also grey-scale (the DEC system had a greyscale display). The latter looks to be a big problem whereas the landscape display should be quite simple both in terms of hardware and software.

Jim Marshall has cancelled all defined delivery dates for products until he has had time to do a full evaluation himself. Once that is done, they will all be redefined with realistic dates that can be met. It was a breath of fresh air!

Brian Rosen has been persuaded to put support for a landscape display in the TEACUP video chip. Apparently it was Brian who was heavily against landscape displays.

The discussion with Jim Gay was for about an hour. Main points:

1. 3RCC are committed to using the SPICE Kernel as the basis for their operating system developments. They will recode D6 in terms of the SPICE kernel initially. Available end of summer.
2. 3RCC are prepared to sell us CANON laser printers. The device will come with the PERQ interface, screen dumper and file printer. Jim Gay will write giving the maintenance terms on which they are prepared to do it. The printer goes directly on the PERQ I/O bus and there would be no difficulty in having it as a print server on a Cambridge Ring - does not touch the GPIB. Can Rob contact Jim Gay and remind him?
3. 3RCC have purchased boxes from LOCAL DATA Corporation called DATA LINKS which attach to the RS232 interface and cause PERQ to look like an IBM 2780/3780 screen. The box is advertised in Computer World. Cost around $2,000. 3RCC would be prepared to sell us one for evaluation.
4. Our ethernet connection was shipped last Friday - probably to the Embassy!
5. 3COM manufacture a box which will connect PERQ ethernet to VAX UNIBUS. Cost around $4,000. Uses IPTCP software and protocols in VAX. 3RCC would be prepared to sell us one for evaluation. 3RCC software for 3COM box should be available early April.
6. Main problem with 16K WCS is that it requires CPU changes and so 3RCC would prefer only to bring it out with KRISTMAS system. However, CMU have a wire wrap 16K WCS which fits on current system and could be produced in small quantities. PCB version likely to be available June or July. Upgrading to 16K WCS means replacing whole CPU board and it could be expensive.
7. I managed to get a specification of the Trust Library. They have not converted everything to D6 yet. If there is anything we need urgently - contact Miles Borel.
8. New tablet is A4 size, very thin, therefore close to desk with a 3-button mouse. Did not see it although I asked twice!
9. There is a GPIB - Diablo interface.

3. CMU

Visit to CMU was disrupted by Three Rivers deciding it was a good idea for me to go Monday afternoon rather than Tuesday morning which meant that CMU were getting two conflicting stories. However, it turned out ok in the end as Peter Hibbard was not around on Tuesday. Also Rick Rashid had not told anybody we need a new version of SPICE kernel. Consequently, I was able to organise it all late Monday and pick up the tape etc on Tuesday before going to the airport.

Main points were:

1. SPICE Kernel has improved since 9 Feb but is in an intermediate state due to Rick Rashid being out of town. Consequently, there may be one or two bugs introduced as well as many being solved.
2. George Robertson has nearly finished the loader.
3. They hope to modify POS PASCAL to run by 15 April.
4. The presentation at the Affiliates Meeting should be of the order of an hour. In order of importance, they would like the talk, a demo, audio visual aids and a write-up. Quite happy even if only first is achieved.
5. Letter giving details of Affiliates and how to join was never sent. Peter Hibbard will make sure it is.
6. CMU are getting worried about the need to belong to the Affiliates in order to use the software. They had felt that they wanted to retain the copyright but provide free licences to use it so that they could at least verify that bug-free versions of systems were being used.
7. They have had a problem with DEC and LISP. Lawyers at DEC have insisted on CMU showing that the software is theirs to give even though it is free. That is, CMU would be responsible for any law suites. As a result, CMU may well put all the software in the public domain. They have a Directors Meeting to discuss this. The ICL position is similar. The purpose of the Affiliates was to bring people together who wanted to trade research ideas and not to be a condition on getting software.
8. Hibbard agreed to bring this up at the Directors Meeting. His view was that if ICL formally asked to use SPICE Kernel and made it clear that they were not going to make money out of it, they would be allowed to have it irrespective of whether they became Affiliates. Hibbard would like ICL to join because they have a lot of common interests particularly in the area of VOICE I/O. Rob should contact Peter Hibbard in two weeks time.
9. Rick Rashid does not need to get Hibbard's permission to visit Lab for an extended period. However, he felt that May would be too early for him to go to UK. July would be better. Term would have ended by then and the hiring of graduate students for the summer would have been completed.
10. Spent a long time talking to Roger Danneberg about a music project on PERQ. Main piece of news from this was that Alan Kay now works for Atari who intend to manufacture a Dynabook. Alan is only consulting at the French World Centre.
11. I picked up a D6 version of DP program from Dario Guise.
12. I have a new copy of SPICE Kernel on magnetic tape from Mary Thompson.

4 CONCLUSION

The trip was worthwhile. It was fortunate that 3RCC had just opened its new plant. A visit later on when Jim Marshall has sorted out what he intends to do would be useful.

On a humorous note, I went into one of the small shops near the old railway station in Pittsburgh. The cashier was moaning because one till was out of action and the other one was on the blink. They were both made by ICL!