freebsd-current Digest, Vol 143, Issue 2

[Matt Dawson]

On Tuesday 02 May 2006 13:00, freebsd-current-request at freebsd.org wrote:
> I tend to get snippy towards the end of release cycles, and I apologize
> to those I've offended or have been needlessly rude to.

You were quite justified, IMHO, and didn't come across as rude at all. Once a 
branch reaches beta or release candidate, the focus is, and always should be, 
on testing existing features, not adding new ones. OK, the RocketRAID driver 
was added for RC2 but that's a one-off safe addition that appears not to 
affect any other stable code, unlike messing with the build tools at this 
late stage. Being trapped between pet feature proponents and folks with a 
timetable cannot be fun and your restraint is admirable, especially 
considering the already high workload on releng at the moment.

Personally, I would not like to see a situation like 5.2 where creature feep 
and not enough testing made 5.2.1 necessary soon afterwards. The releng team 
got the blame but it was the demanding pet feature owners that caused the 
problem, AFAICS. Such things reflect badly on the project as a whole, unlike 
not being able to view the WMV funny sent by your co-worker from two cubes 
away, which is trivial by comparison and only reflects on the application's 
inability to be 64bit clean or being a closed-source binary blob which is 
only compiled against i386.

BTW, running RC2/amd64 and so far it seems rock solid and very fast (empirical 
evidence only - I don't do synthetic benchmarks as they tend to tempt you to 
add silly things like -ffast-math to make.conf in a vain attempt to get 
similar performance to whatever you're testing against on a workload that is 
rarely likely to crop up in Real Life [TM]). I'm having a few problems with 
Xorg and a Radeon 9700 on one desktop, but that is nothing to do with the 
base system.

My opinion only, of course.
-- 
Matt Dawson.

matt at mattsnetwork.co.uk
MTD15-RIPE OpenNIC M_D9
MD51-6BONE