Re: Deferred reads

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Thu, 01 Jun 2000 16:44:18 +0200

Duane Wessels wrote:
>
> On Wed, 31 May 2000, Andres Kroonmaa wrote:
>
> > actually, squid used to skip defered FD's from being polled, which
>
> Yes, thats the whole point of the deferred read handlers.

We do know. However there are other more efficient means to acheive that
than CPU polling a lot of defer handlers on every poll loop.

You and Adrian are talking by each other, one only looking at how the
current code operates, and one forgetting about the constrains of the
current code and struggling to to something completely different..

Adrian, the trial and error method probably won't get you where you want
here, at least not in a timely fashion. The event driven model needs to
be carefully designed, possibly prototyped in a smaller scale and then
implemented. Simply jumping into the code and trying to kill the
deferers is not a good approach I think. The current design heavily
depends on the defereres.

/Henrik
Received on Thu Jun 01 2000 - 08:51:59 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:27 MST