 |
|
||
|
|
OOPSLA 1999 trip report, Thursday |
On Thursday morning, which was early given the five hours sleep per night I was getting came much too early.
The highlight of the day was Dan Ingalls Squeak Talk. 'Fun with Squeak and Other Smalltalks'. Now remember these are professional programmers they've seen it all, but Dan blew their socks off.
We've all seen Squeak, sure we know it can do video, it can do music. But we forget it's all done in Squeak, no OS routines or third party software involved, no QuickTime, no magic, just Smalltalk. Great but it's wonderful when you add content. Dan first showed how to take some existing Disney video and concert music, link them together and sync the timelines so the music scores matched the tempo of the video by dragging and dropping video frames into the music score. Most impressive.
Smalltalk-72. It lives. For Alan Kay's birthday the Disney folks scoured the bowels of Silicon Valley looking for an Alto, and they found a couple of old hard drives. What would you do with a hard drive that was created 28 years ago? Plug it into your PC? Well I'm not sure how they did it but I was informed later that lots of documentation exists on the hardware so it wasn't a problem, and they were pleased to find on the disk the interpreter and lots of historical software. Dan did admit the undocumented font files were a problem for awhile. Armed with these bits they casually created an Smalltalk-72 image within a Squeak image and proceeded to show us how it worked, neat. I hope they release it since it's quite different when compared to Smalltalk-80. {PS Done. Note the Smalltalk-72 link above which links to Dan's note about this}
Dan went on to show Alice which got lots of oohs, and then he started up Etoys and showed how scripting worked. From there he showed a set of examples with simple to extraordinary content. The use of a magnifying glass, a morphic construction helped to show the detail.
For example a set of gears, where the physic of a gear tooth within the script is showing how the mechanical effort of a gear tooth pushing another tooth can turn the other gear. Or perhaps Winnie-the-Pooh and Piglet dancing over a bridge. (Yes Disney Content) In this example Winnie-the-Pooh was used to show Balloon and Flash where a bear was 200K as a bit image, but in Flash he became a 1K object. Of course Dan picked up Poo Bear and deposited him in a corner and examined the detail with the magnifying glass, more oohs and awhs.
Well this was going on, I and other Smalltalkers felt very insignificant. For some of these problem domains I'm sure it would take three months of reading to understand and months of coding, then I'm not sure if it would be as good as what I saw. But remember who these folks are.
As for Etoys, later Dr. Johnson told me we was going to unleash his kids on it and later they could teach him how it worked since we are boxed in by our views of how it works and our worry about how the Smalltalk under the covers is doing the magic. Round corners versus square windows perhaps.
Dan did show off some tools we could understand from a programming viewpoint. Like Ted Kaehler's selection browser. In this browser selector you enter input and the expected output values. Say 3 4 7. IE. you input 3 and 4 and expect to get seven back. The tool went away and of course presented 3+4. But it also presented other methods like 3 bitXor: 4
Very neat, but on Friday at lunch, Don Roberts told me he ported it to VisualWorks and showed how to break the image by using ObjectMemory as the target class versus Magnitude. Beware of side effects eh?
Later in the day I got to sit with Andreas Raab and Ian Piumarta as Ian worked on a method caching system to cache JIT opcodes on a method basis. After some coding problems we had success with a system that provide a 94% cache hit solution. While this was happening on the other side of the table Andreas was moving pieces of BitBlt to "JITBlt" an external plugin that used Ian's pseudo assembler to increase performance. This was more difficult problem since Andreas was being exposed to Ian's assembler without documentation but at the end Andreas got a 2.6 increase as expected in performance for one of his BitBlt test cases, we were all pleased.
At one point I got a cappuccino from the adjacent coffee bar and got talking to a fellow who was excited by the Squeak work and I remarked to his amazement that over there was Ian working on JIT, and Andreas working on BitBlt and I believe Craig Latta was working on Squeak at another table. He was amazed there within 8 meters was serious ground shaking Squeak development work happening.
{Thursday afternoon my WSJ alert email announced the purchase of ObjectShare by Starbase, much confusion Thursday and Friday about what in the world StarBase thought they were buying, but the people holding ObjectShare/Parcplace/Digitalk stock options were ecstatic!}
Much later in the evening after the "Wings over the Rockies" show we retired with a large group of people, mostly Smalltalkers to discuss the days events and of course Smalltalk. I believe it was John Sarkela of "The 4th Estate" now ExoBox who in our discussion about the use of Smalltalk versus Java hype coined the analogy from The Wizard of Oz. "Hey ignore those Smalltalkers behind the curtain just watch the flash, smoke and mirrors from this Java implementation, after all who is doing the real work here anyways"