How is Intel punished? [LWN.net]

How is Intel punished?

Posted Sep 5, 2007 22:18 UTC (Wed) by khim (subscriber, #9252)
In reply to: How is Intel punished? by epithumia
Parent article: LCE: Linux, hardware vendors, and enterprise distributors

To develop driver for new kernel and to develop binary blob for ancient kernel in some "enterprise distribution" is to spend more-or-less equal effort: if first place you need to work with community (and it requires time), in the second - you need to fight bugs/shortcomings of old kernel. The proper driver for new kernel is not easily adaptable to old kernel - so in effect Intel is punished for support of open-source development model. I see no easy choice except one: treat the whole kernel and HAL level and upgrade it over time - even in enterprise distributions. Of course it requires testing - but it's needed anyway because "old kernels" in enterprise distributions contain tons of backports...

(Log in to post comments)

Vendor-distributor cooperation

Posted Sep 11, 2007 19:36 UTC (Tue) by hazelsct (subscriber, #3659) [Link]

The obvious solution is to work together with the enterprise distributor. For example, provide a driver which works with the latest kernel, and split the cost of a backport with the distributor(s). For Ubuntu Dapper, the kernel is 2.6.16, and a long series of 2.6.16.n releases have had various fixes and backported drivers.

This in turn would create an incentive for distributors to coordinate their enterprise kernel releases, in order to reduce everybody's backporting costs. This idea isn't new, Ubuntu developers have been requesting it for some time...

Of course, massive API changes such as mac82011 wreak havoc with such a process, but they don't happen with every single kernel release.