Skiplist / best practice for 2.1 branch

[Henrique de Moraes Holschuh]

On Mon, 18 Nov 2002, Lawrence Greenfield wrote:
> Some of them (the lock timeout/wait is what I have in mind) will make the 
> system perform not as well.

I'd have to profile it to know more, and I am quite short on time to do that
right now.  Maybe the fastmail.fm crew has any comments on this issue?

> (we suck, sorry), or it feels very OS specific. So a lot of it is mostly 
> waiting on me.
> 
> Hopefully if something is really painful for Henrique to maintain 
> seperately, he'll tell cyrus-bugs and we'll work it out.

Well, there are two sets of patches I really would like to make it upstream.

The first one are the shutdown() ones.  They're quite safe, and they could
increase wire performance in other kernels as well.  And it IS a bother to
periodically go all over the code trying to track down new calls to close
that need a shutdown().  And I am sure I missed a few places where one side
of the conection could be shutdown earlier due to the protocol...

The second one is the master pidfile control.  It is very safe, useful, and
portable (the code was borrowed from Vixie cron...).  It also makes for far
better behaved init scripts.

Both are simple patches, if a bit scattered all over in the case of the
shutdown patch.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh