[Moved here from [EMAIL PROTECTED] On Thu, 6 Jun 2002, Justin Erenkrantz wrote:
> On Thu, Jun 06, 2002 at 04:51:10PM -0700, Aaron Bannert wrote: > > I'd *love* to see progress on something like this, but whenver I start > > thinking about this kind of a MPM I run into the problem where brigades > > (or filter chains) can't be multiplexed over. Is this what you have > > in mind? (Let's get a discussion going.) > > Haven't we talked about adding an apr_bucket_poll function? I > don't think I saw any objections to that the previous two times > we've discussed it, so I guess it's waiting for someone to come > along and do it. -- justin I've objected to it off and on ... in particular forms. I just don't see what good apr_bucket_poll would be if you can only poll one bucket, which is what you'd have to do to be consistent with the other apr_bucket_foo() functions' semantics. Polling one bucket at a time is no different than doing a nonblocking read on each one in a tight loop. It's the tight loop obviously that's the problem in either case. So what we really would need is some way to add a bucket (or a brigade) to a "pollset" of sorts. In either case it means that we would need a way to add individual buckets to a pollset. But that kind of depends on having poll() or select() or whatever being able to interact with all of the kinds of buckets from which reads might block, and that might not be possible. I'm kind of just thinking out loud here, but I do think this is easier said than done... --Cliff