httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Trawick <trawi...@bellsouth.net>
Subject Re: [PATCH] APRize nagle
Date Mon, 21 Aug 2000 16:57:16 GMT
rbb@covalent.net writes:

> On 21 Aug 2000, Jeff Trawick wrote:
> 
> > "Brian Havard" <brianh@kheldar.apana.org.au> writes:
> > 
> > > The current ap_sock_disable_nagle() requires an OS handle. This patch makes

> > > it take an apr_socket_t * instead. Note that I had to stick my fingers into

> > > MPMs & APR code I can't test which is why I'm posting the patch before

> > > commiting it. Is this ok by everyone or do we need to keep the FD_SETSIZE test
> > > in there? I removed them because this removes the last use of the os socket
in
> > > that part of the code.
> > 
> > We need the FD_SETSIZE check *somewhere* (APR fails on certain
> > operations otherwise).  I'd leave that code as-is until/unless
> > somebody changes where the test is done or removes the need for it.
> 
> Couldn't we move the FD_SETSIZE inside of APR?  

Sure...

>                                                  I am envisioning something
> like:
> 
> apr_disablenagle(apr_socket_t *s)
> {
> 
>     FD_SETSIZE magic
>     setsockopt magic
> }
> 

This isn't the right interface.  FD_SETSIZE has nothing to do with 
Nagle.

apr_create_tcp_socket() and apr_accept() could compare new descriptors
with FD_SETSIZE, realize that APR is broken (when they're too big),
close the socket, and return a special error code that causes the app
(e.g., Apache) to issue a trace that the clever admin can decipher.

-- 
Jeff Trawick | trawick@ibm.net | PGP public key at web site:
     http://www.geocities.com/SiliconValley/Park/9289/
          Born in Roswell... married an alien...

Mime
View raw message