Chairman’s Impressions: What you Saw was (more or less) what you Got

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 42

After a lapse of a couple of months, it is interesting to reflect on what we set out to do at Swansea, and how close we came to achieving it. Of course the objectives were never totally stable — we were trying to do something quite new — but there were some fixed points in our agenda, and it is reasonable to ask:

• did we really show you the newest and best work being done under Windows on a PC?
• did you pick up any leading edge techniques of quantitative management?
• were you able to spend the afternoons taking apart the morning plenaries in a well-equipped laboratory full of fast PCs?
• was it a revitalising break from the daily routine?
• did you get to take away some useful code?

These were all promises that we made on the first page of the invitation: how well do you think we met them? So far, those of you who have written or e-mailed have all been remarkably positive (which makes me feel better), but such samples tend to self-select! I might yet get around to writing to everyone who came, to ask how we could do it better next time, but meanwhile, here are my views on what we did well, what we did badly, and what we left undone.

The Venue, Format and Organisation¶

Swansea was born out of our failed bid to run APL94 at York, and for the much smaller workshop format, I think it was close to ideal. We began by over-estimating the size of the gathering, which caused quite a few headaches as we cut back to match the bookings, but it would have been hard to run such an intensive 3-day workshop with many more people.

The EBMS was a good self-contained base, the only snag being the need to borrow facilities in the nearby computer lab for some of the more heavy-duty afternoon sessions. However most people found their way around, and in the end most of the equipment co-operated, so this somewhat ad hoc approach stood the test — just.

Several people commented favourably on the basic structure of running a morning plenary, with everyone attending all the sessions. This had the great advantage of simplifying administration, as you could always be sure that everyone had got the message about the bus times, or the change of room in the afternoon. I hope the more esoteric statistical stuff didn’t cause the old-time APL hacks too much mental indigestion — I thought Derek Bissell’s talk on Control Charts was brilliant, but then again I learned my statistics by weighing Kit Kat bars! I honestly think that if we had gone for a straight-down-the-line Windows + GUIs agenda, we would have turned everyone off by the end of Day-1. It needed breaking up a bit, and talks like Kath Dowsland’s on Genetic Algorithms were ideal.

I can take no credit at all for the excellence of the organisation! This was all set up by the Alans (Sykes and Mayer) and administered most capably by Haydn Williams and Pam. The coffee was on time, the signposting was pretty good, the overhead projector worked all the time — what more can you ask?!

The Main Topics¶

We took a quick poll around the room at the start of the first day (see Jon Sandles’ notes on page 45), and most people had really come for the experience of Windows programming. We had already warned everyone (Vector 10.4) that this was not a jolly, and I think we went home feeling mentally quite battered by the amount of new stuff we had been exposed to in such a short time.

• Day-1 was short of ‘talk and chalk’ time for the main topic of programming the GUI APLs, so the afternoon workshop was something of a rabble. I had failed to appreciate the problems of class-control when you take 24 keen APLers, sit them down in pairs in front of a new toy, and try to get them to listen to you. In the end, Duncan and I gave up any attempt at formal teaching and just worked around the various groups as fast as we could. It was OK, but we did much better on Day-3 (see Ray Cannon’s note on the DLL workshop) when we prepared more material in advance.
• Day-2 was about client-server, timeseries and graphics. I found Morten Kromberg’s lecture on the basics of effective client-server architectures really superb, and I’m sorry I missed the afternoon workshop. The idea of a room full of chunky 486 ‘clients’ all happily running database queries against Morten’s dinky little SparcStation ‘server’ was remarkable enough. What’s more it all worked!
• Day-3 was initially scheduled for Microsoft, who are anxious to tell the world about OLE2. In the end, we were too small a gathering for them, and instead we had John Scholes on Namespaces in Dyalog-7. I think that for many people this was actually a bonus, and namespaces were a hot topic for the rest of the conference. In the afternoon we had a much better organised workshop on DLLs; we need to remember how we did this one and use it as the pattern when we do this kind of thing again.
• by Day-4 everyone was pretty shattered, so we had a gentle 2-hour canter through some simple GUI applications, made sure everyone got the most up-to-date software (the workshops turned over quite a few bugs) and unwound gently ready for the trip home.

To summarise, I think that we delivered most of what was promised, but I was too deeply buried in running the thing to get a clear view. Let us know what you think, and the BAA may try to do it all again, better.