On Sun, 2009-05-03 at 19:34 +0100, Matthew Garrett wrote: > On Sun, May 03, 2009 at 12:07:43PM -0500, Callum Lerwick wrote: > > On Fri, 2009-05-01 at 00:48 +0100, Matthew Garrett wrote: > > > The problem with the desktop is that it's obvious, looks simple and > > > everyone has an opinion on it. But not all of these opinions are equally > > > valid. This kind of situation is much easier to deal with in, say, the > > > kernel VM system - in that case it would be perfectly acceptable for > > > people who spend their entire working lives concentrating on a specific > > > topic to say that they know better than people who occasionally touch > > > upon it. > > > > ... And the kernel VM people tend to be able to back up their expertise > > with hard data. If a patch isn't an improvement in some provable manner, > > it doesn't get in to the kernel. > > Like everything else, kernel VM performance is a tradeoff. Improvements > may benefit certain workloads while impairing others. It's rare for > changes to improve things for everybody - so the end result is a > judgement call, generally by the people who are assumed to know what > they're doing. No, it is not a judgement call. Hard performance measurements are not subjective. We developers are not sheep, and this is not a closed proprietary project. We expect our experts to show their work.
Description: This is a digitally signed message part