Keeping a single PC up to scratch these days can be hard work. Assuming it is working properly out of the packaging, it is only a matter of time before some piece of new software uncovers an old bug in existing hardware or drivers. The right patch has to be found and applied and, given the complexity of the modern PC environment, checked against whatever else is in the machine. Microsoft has gone some way to easing this process with its auto-update scheme, but it is not uncommon to lose hours of work finding and fixing one small latent problem because of other, seemingly unconnected, work.
Now imagine having a fleet of hundreds of laptops and desktops, each of which may have different levels of hardware, driver or software upgrade present and each of which may stop working if an inappropriate patch is fitted. Even the most assiduously standardising IT manager can't buy all of his equipment at exactly the same time. With chip and subsystem manufacturers committed to many small upgrades over the lifetime of a product, two apparently identical computers bought a month apart may have subtle but important differences in their software image -- the snapshot of the specific OS, driver and application mix that runs on a particular computer.
The problem is amplified when computers go wrong and are repaired or upgraded with whatever's current at the time of failure. Soon enough, a multiplicity of software images are needed to match the many small variations, with no guarantee that a vendor will continue to support the older ones. It all equates to a great deal of money and time wasted, and a considerable disincentive for IT departments to upgrade after their particular maze of twisty little problems finally reaches stability.
In response to this problem, OEMs have been pressing chip and subsystem vendors to guarantee fewer changes and more stability than before. Intel has been receptive to this idea, initially announcing a rather ill-defined initiative called Granite Peak at the Spring 2003 Intel Developer Forum. This has now been fleshed out and presented to the world as the Stable Image Platform Program (SIPP).
Granite Peak's promise was to limit driver revisions for new Intel chipsets for a period of six months, making the hardware platform much more consistent for programmers to write for and system deployers to support. SIPP goes further, saying that revisions will be limited for a full year.
The initiative applies to a subset of Intel's desktop and mobile products, and is tied to the launch of major chipsets. On the desktop, the Intel 865G launch in May 2003 marked the beginning of the SIPP period for system driver software for that component -- including LAN and integrated graphics drivers. The first mobile SIPP provisions apply to the Centrino family, based around the Intel 855 -- LAN, integrated graphics and wireless are included here, and the period is retrospectively deemed to have started in March 2003. To help companies plan their upgrades to coincide with SIPP periods for new systems, Intel has paid more attention to roadmaps, simplifying and amplifying them in line with the initiative: new products have a qualifying and launch phase during which, one assumes, initial problems are identified and rectified. After that, the guaranteed period of stable deployment starts.
Intel's announcement quotes analysts at the Meta Group, who say that more consistent PC platforms can save large companies with complex PC infrastructures up to 15 percent on IT support and management. The company is also trying to get others on board, saying that as the leading manufacturer of many core PC components it is in a unique position to persuade others to coordinate component life cycles, driver revisions and manufacturing schedules. Several major PC manufacturers are already supporting SIPP, including Dell, Fujitsu, Gateway, HP Compaq, IBM, and NEC.
There are limits to how much efforts like this will ameliorate image multiplication madness. Serious problems coming to light after launch will always have to be fixed, while it seems unlikely that the constant stream of essential security patches covering operating system and applications software will dry up any time soon. It's also not clear how many companies will be willing or able to synchronise their update cycles with Intel's product plans, nor whether the channel will be happy with the boom-and-bust sales pattern that will come about if such synchronisation works: it'll be the August new car registration syndrome all over again.
Nonetheless, SIPP has met with general approval and it looks likely that it will ease the lives of corporate IT departments with little downside. If it's the forerunner of a period of greater stability within the industry as a whole -- with big changes limited to genuine innovations rather than an endless stream of tweaks -- then it will signify a welcome maturity.