Mailing-List: contact new-httpd-help@apache.org; run by ezmlm
Precedence: bulk
Reply-To: new-httpd@apache.org
Sender: gregames@raleigh.ibm.com
Message-ID: <3A5CD267.373D368A@raleigh.ibm.com>
Date: Wed, 10 Jan 2001 16:21:43 -0500
From: Greg Ames <gregames@raleigh.ibm.com>
MIME-Version: 1.0
To: new-httpd@apache.org
CC: rasmus@apache.org
Subject: Re: woah, "GET /" with autoindex
References: <Pine.LNX.4.21.0101071329540.17843-100000@koj>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

rbb@covalent.net wrote:
> 
> > > > since the underlying ap_bucket_putstr and ap_bucket_printf are implemented
> > > > as one-copy, coalesce will result in a two-copy implementation... which is
> > > > less than 1.3's one-copy implementation.
> > > >
> > > > not to mention that the current ap_bucket_{putstr,printf} result in far
> > > > too many ap_bucket_ts being allocated.

[....]
 
> They are close, but they aren't exactly the same.  If module is using the
> ap_r* functions, then we can assume that it isn't using the buckets
> directly, and we can buffer.  

Didn't Victor post a patch not too long ago that did exactly this?  IIRC
it was vetoed. Victor, if I'm not imagining things, could you please
re-post to save everybody some time?

>                             If a module is using the buckets directly,
> then where do we put the buffer?

What buffer?  I thought we were talking about lots of itty bitty data,
like dynamic html.

Assuming we fix ap_r*, are there any modules out there now that generate
tons of itty bitty data directly to buckets, or any modules likely to do
it? 

(I would think you would go directly to buckets with big data - files,
mmaps, probably pipes - where you typically wouldn't use a buffer above
the kernel.)

If I'm correct, what would a new buffer in the buckets path buy us?

Rasmus, which API do you see mod_php using?  Does PHP buffer internally?

Greg