Quacking the SPARC barrier

a demonstration of the strategic value of giving away "crown jewel" class information
Written by Paul Murphy, Contributor
Not so long ago, and in a world near you, the BSD people had a knock down, drag-em-out public relations fight with Sun over access to the memory management and related hardware details on the UltraSPARC2. Prior to the 200Mhz release that machine had a bug, and some people at Sun were sure that a couple of noted BSD contributors were behind the software that exploited this to cause some machines to simply stop.

Some Linux releases for SPARC suffered from the fallout too, mostly by simple geographic association, but enough NDAs got approved in the U.S. that a few SPARC ports continued, although with a lot less fire and enthusiasm than before.

Meanwhile the world changed: Sun brought out the first products in its bet-the-company SMP on a chip line, released the full Hypervisor API specifications for it in February 2006, and open sourced the entire hardware specification in mid March.

Three months later we'll have "Dapper Drake", the next generation Ubantu release (based on Debian Linux) with full commercial support and specific adaptations for the Niagara version one CMT processor.

I believe that's the fastest Linux has ever gone from generic to a commercially supported, hardware specific, release on a brand new architecture. Even by internet standards three months is just amazing -and sure, it's a tribute to the strength of the Linux community, but I think it will ultimately be remembered more as a demonstration of the strategic value of giving away "crown jewel" class information.

Basically, hardly anyone noticed what Sun was doing at the time and those who did generally saw Sun giving away critical information for nothing, but now the Linux community gets invited into the SPARC/CMT camp - and Sun gets another opportunity to win developer loyalty as the battle to replace x86 heats up.

Editorial standards