SPARC: The black-eye of open standards

Right now, I'm cursing both ZDNet's search facility and Google because neither are turning up a story that I wrote several eons ago about SPARC, Intel x86 and the definition of a real standard [Update 6/10/2005: ZDNet reader Brian Green found it].  I have Sun CTO Greg Papadopoulos to thank for sending me on the wild goose chase.

Right now, I'm cursing both ZDNet's search facility and Google because neither are turning up a story that I wrote several eons ago about SPARC, Intel x86 and the definition of a real standard [Update 6/10/2005: ZDNet reader Brian Green found it].  I have Sun CTO Greg Papadopoulos to thank for sending me on the wild goose chase.  In his most recent blog (triggered by he Apple/Intel news), Papadopoulos shrouds one of the worst black eyes for open standards in an appeal to Intel to open up the x86 architecture so that others (be they virtual machine or silicon providers) are free to come up with their own implementations.  Well, not exactly implementations of x86, but rather, implementations of a specification he calls the open core instruction set architecture (ISA).  Presumably, given Intel's footprint in the market, this ISA would be predominantly colored with x86 or AMD64 instructions while the architecture's open nature would pave the way for the entire industry to collaborate over the evolution of a single universal microprocessor architecture.

The hook to the Apple/Intel deal has to do with the expense that, in addition to the OS vendor, the software developers must bear to maintain multiple codebases.  There's more to maintaining multiple codebases than Apple probably realizes, implies Papadopoulos.  And who should know better than Sun, which is currently engaged in precisely the same practice for Solaris.

Microsoft may be able to shoulder that cost for the OS X version of Office, but it creates a burden for smaller developers and regardless of a developers' size, the cost must ultimately get passed down to us buyers. So,I'm in fervent agreement with Papadopoulos.  What better way to drive the cost out of computing than to drive a complete waste of a cost like that (one that could be ameliorated by something like an ISA) out of our hardware infrastructure.  Not to mention driving out the premium that Intel and AMD get to charge for their chips because of the proprietary nature of the instruction set -- a premium that, theoretically, would collapse under the weight of a free market with multiple competitors (Intel is in triage as a result of AMD, but it could be far worse).  Not surprisingly, Papadopoulos doesn't give much treatment to how his proposal would affect Sun's blood-brother AMD.  As long as x86 continues it's dominance, AMD's and Intel's futures are inextricably linked by a legacy cross-licensing tryst that both companies would just as soon be free of.  It's not a perfect world. Opening x86 up would work to AMD's favor since it has demonstrated just how easy it is to come up with better implementations than Intel of the same architecture.  It would also probably cut Intel off from any access it has to the IP of AMD or any other "ISA" competitor.  While users, software vendors, and would be Intel-competitors have the most to gain, those gains would be Intel's loss.   Hell (or Intel's profits) will have to freeze over before Intel opens up. 

Unfortunately, there's no clear evidence that such an open architecture would win anyway.  History (and in his blog, Papadopoulos) tells us that when the industry had access to such an open  standard -- SPARC -- it had very few takers.  Surely, after all these years, if the market demanded an open processor architecture over a proprietary one, SPARC would not only have tamed x86, but would have also rendered IA-64 (inside Intel's Itanium) and AMD64 (inside AMD and Intel's 32/64-bit hybrids) dead on arrival.  Sadly, despite its open nature (especially when compared to Intel), it never did. In the spirit of recommending open standards over anything proprietary (and the way that puts buyers instead of vendors in control), I have always rallied for the underdog SPARC.  Fellow ZDNet blogger Paul Murphy has been outspoken as of late too here, here, and here. Not to mention Sun COO and president Jonathan Schwartz, who pulled out his trusty (or is that rusty?) ole SPARC rhetoric for the Apple/Intel news.  But let's face it: SPARC is the black-eye of open standards.  It's proof that sometimes, proprietary wins.

Sure, we can blame the Wintel marriage for assuring x86's destiny and talk about how SPARC was often maligned as proprietary while x86 was hailed as the standard when in reality, it was really the other way around. But there's plenty of evidence to suggest that when a technology's value proposition is out of whack with the market's needs, the market will move to a new "standard," open or not.  Linux certainly proved this. So too did Apache. (Neither one is an open standard.)  TCP/IP has prevailed over just about everything Microsoft, Novell, Banyan, and others could come up with and look at what's happened with HTTP and XML (although all may not be as it seems in XML-land, a topic for another blog).  For the time being, boutique RISC technologies will have their place in the market.  Just ask the thousands of customers who giggle at the idea of transitioning from their SPARC-servers to something Intel-based (eg: CEO Marc Benioff, a man who had the luxury that very few do -- picking any technology to run his business).  If a standard CPU architecture prevails, it will only be after Intel and AMD fail to serve the market; and, as far as I can tell, that won't be any time soon. 

Finally, as a side note, I'm beginning to wonder whether Sun executives actually get blogs or not.  Sun COO and president Jonathan Schwartz gets a lot of coverage for being a high profile executive blogger (if not the  highest).  But given the way his blogs don't accept comments, he's only broadcasting as opposed to starting or engaging in conversations (which is what I thought the blogosphere is all about).  Not only can't you comment on Papadopoulos' blog either, he only broadcasts about once per month.  That infrequency is apparently cause for concern to Sun's public relations folks since, every time he blogs, they feel the need to remind the press  of the momentous event via email.  As though RSS isn't enough once we learned that he was blogging, er, broadcasting.  So, Carrie, thanks but no thanks.  I'm subscribed to Papadopoulos' feed.  You can take me off your bcc: distro list.