On Thu, Jun 09, 2005 at 12:19:09AM +0900, SODA Noriyuki wrote:
> >>>>> On Sun, 5 Jun 2005 14:01:57 +0900,
> 	Izumi Tsutsui <tsutsui@ceres.dti.ne.jp> said:
> 
> > Should we also add #options BUFQ_PRIOCSCAN in these files?
> 
> Yes, I think so.
> 
> As Jason said, BUFQ_PRIOCSCAN is considered better than BUFQ_READPRIO.
> So, why don't we add it to where there is already BUFQ_READPRIO?

If SGI's claims can be believed, BUFQ_PRIOCSCAN should do better both
on clients and servers than either the traditional sorting (which is
pretty much CSCAN) or pure read priority.  I would support making it
the default -- we can change it back before the release if problems
turn up.

An interesting reference that gives a few other policies that might
not be too hard to implement is http://portal.acm.org/citation.cfm?id=511373

-- 
 Thor Lancelot Simon	                                      tls@rek.tjls.com

"The inconsistency is startling, though admittedly, if consistency is to be
 abandoned or transcended, there is no problem."		- Noam Chomsky