Jon Udell has an interesting session with Steve Burbeck (who's only apparent career blemish appears to be the design of UDDI 8^). [Update:
Burbeck's history goes back to that Tektronix Smalltalk community here in Portland that spawned Ward Cunningham, Kent Beck, Rebecca Wirfs-Brock, Gemstone Smalltalk, Digitalk Smalltalk, etc. The work Cunningham, Burbeck, and others did at Wyatt with a Smalltalk-based trading system is a gold mine of ideas barely tapped, yet more relevant than most enterprisey systems built since.
Someone could write a book on Tek Labs and the direct and indirect influence it had not just in popularizing Smalltalk, but design patterns, agile programming, etc. It's reach is far and wide and largely unrecognized.
Back to Burbeck... the current references are to his work on multi-cellular computing. Apoptosis is an increasingly recorgnized pattern, e.g. from the Erlang community the idea that small components should give up quickly and allow a higher-order component handle the fault.
Stigmergy is interesting... it's not really found in Erlang, except I guess it is in a sense what the Mnesia distributed database is for. It's also a key aspect of the Linda tuple spaces / Javaspaces model, and the idea of "blackboards" in artificial intelligence.
3 comments:
First, in my defense, I was only one of the architects of UDDI v1. That first version was already getting too complex but was nothing like as bloated as the later versions. My primary contributions were around the taxonomies (something like I described in the "Tao of e-business services" paper). Microsoft didn't want any taxonomies. They envisioned UDDI as whitepages only, which in my view would have been totally brain dead. Steve Graham and I fought for taxonomies and finally prevailed. But we were discouraged by the complexity phreaque's domination of UDDI and bailed after v1.
On the apoptosis pattern...I wasn't aware of the Erlang parallels. Neat. I'll look into it.
I like to think of the four multicellular principles as interdependent. The situation is somewhat like Asimoves "Three Laws" of robotics. If you ever read any of Asimov's robot novels, you'll remember how it was the subtle interactions between the laws that made for interesting cases. All four of the principles I've highlighted evolved together in biological systems. My guess is that multicellular computing will need all four too.
Steve Burbeck
Is apoptosis somehow supported at the language-level in Erlang? (or did I read too much into the comment?). Stigmergy supported at the langauge-level is an interesting idea.
As far as the legacy of Tektronix goes: It's unfortuneate that we don't have anything similar going on in the Portland area now. Tek was sort of our Xerox PARC back in the 80's and early 90's. While Intel now employs almost as many people in the Portland area as Tek did during it's heyday, it doesn't seem to be having nearly the same kind of impact that Tek had (where 'impact' can be measured in startups, spinoffs, research). Several other companies came out of Tek - Mentor Graphics being a notable example - I don't see the same happening with Intel possibly because Tek was a much more 'open' company. It's also partly due to the fact that there really isn't much 'research' going on anymore at any companies. HP back in the day before it just became a PC OEM company was another great example of research-focused engineering company.
It is not at the Erlang language level per se. Support for it is part of the process management API which is used in nearly every Erlang application.
Post a Comment