Peter Coffee

By eWEEK Labs  |  Posted 2002-11-18 Print this article Print


With all these obvious arguments against it, Itanium is poised to make its critics look like idiots two years from now.

Technology creators often have to make a difficult choice between protecting current users from unreasonably disruptive change or protecting future users from the errors of the past. The anger of customers today is clearly more of a hazard than the fear of derisive comments by prospective customers tomorrow. Its no surprise, therefore, that most of the errors occur in the direction of taking too long to decide that a change is due.

The coming year will be filled with more chances to make this same mistake, and even the best-considered advice may someday be seen as an example of this error. Ill take the chance.

For example, Intel Corp.s Itanium initiative continues to look like a truly bad idea. It relegates the huge installed base of 32-bit x86 code to run in an emulation ghetto, off in the corner of a chip that will devote the rest of its resources to a completely new instruction set.

Further, it hopes that the IT buyers who will need 64-bit power tomorrow havent already needed it enough to have made a commitment to existing 64-bit hardware from IBM, Sun or other providers. It demands the support of a completely new generation of compilers, debuggers, device drivers and other system software. It also seems like a terribly mistimed arrival-these days, enterprise IT is implacably hostile to anything that looks like change for its own sake.

With all these obvious arguments against it, Itanium is poised to make its critics look like idiots two years from now, when it may be that next year will be generally acknowledged as the year when Itanium turned the price/ performance corner to leave Advanced Micro Devices Hammer in the dust. With its 64-bit superset of the x86 instructions, and its road map toward continued performance improvement for 32- and 64-bit code on a common hardware base, the question is whether Hammers intrinsic hardware complexity has the headroom it will need-or if Itaniums gamble on shifting complexity to system software will pay off. Im betting on AMD.

Ill take a similar risk in my other beat, application development technologies. The coming year will mark developers first full year of access to a shipping Visual Studio .Net, and thus to Microsofts Common Language Infrastructure, with its claimed advantages over Java.

At the end of next year, its my expectation that Java will not only be still standing but also still growing in enterprise acceptance. To paraphrase Sun Fellow and Vice President James Gosling, in a conversation that we had earlier this year, sometimes language design is more about making it impossible to do stupid things than about making it possible to do extremely clever (but also quite dangerous) things.

I see more problems arising from software thats stupid than from software thats not as clever as it could be-and thats why Java still looks good to me in 2003.


Submit a Comment

Loading Comments...
Manage your Newsletters: Login   Register My Newsletters

Rocket Fuel