Open Source Hardware?
By webmink on Dec 06, 2005
One of the things that cropped up in the European software patents debate at the start of the year was the issue of defining what a software patent actually is. You'd think it was easy, but it's not. Software patents sneak through by reference to the computers they actually run on (a software patent I have to hand starts "Two computers A and B each have a clipboard function..." - that's a European one, by the way). And even chip designs turn out to be a grey area.
I recently had the opportunity to meet some of the team designing the new UltraSPARC T1 processor ("Niagara") and saw first-hand what my friends up the road at Southampton University have been pointing out for years. While there's a good deal of skill in the instantiation, a silicon chip these days is "just" the compiled version of a software design. A chip like the UltraSPARC T1 is actually a huge Verilog program compiled to atoms instead of assembly language.
The reason I found all this out is that the team has decided it's time for the Participation Age to reach into the world of chip design. Nestling among all the other excellent announcements about low energy, high performance computers in today's launch event is an announcement that to me is quite revolutionary. SPARC is going open source.
Yes, you read that right. The Verilog source code, tools and more behind the UltraSPARC T1 (the "design point") will be released under an OSI-approved open source license next year - OpenSPARC - and a community will hopefully be forming to use that design point for any purpose that's interesting. I think that's revolutionary.
Of course, open source hardware doesn't work the same way open source software does. Because the "compilation" process is so heavy, the community won't be working on a single rapidly-iterated shared tree of source. It's likely to work more like the Jini community, with many co-operating community participants doing their own thing with a common baseline and then contributing back innovations and fixes based on their experience.
I'll be helping as much as I can to get this off the ground, and learning a lot in the process. It's heading into uncharted territory and there's plenty of opportunity to "learn through corrected error" but I think the new OpenSPARC community is the start of something big, the perfect complement to OpenSolaris.