httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: lingering_close performance idea
Date Tue, 11 Feb 1997 01:39:08 GMT
When Dean and I talked about this, I had some reservations about doing
anything with descriptor passing before 1.2.  The problem is that I am not
convinced of its portability.  For example, the NCSA code will only work
with systems that support more than 256 fds/process.  And it needs three
main ways (BSD, SYSV, and Linux).  It would make the cleanup a _lot_ nicer
since it could be handled from the main parent loop.  

It is cleaner than handling it in the child, but I think portability could
be an issue.  I'll have to examine the possibility of handling it in the
child, the effect of lowering the timeout to 1 sec, and the fd passing
stuff in more detail before I can comment further.

On Mon, 10 Feb 1997, Chuck Murcko wrote:

> David Robinson wrote:
> > 
> > On Sun, 9 Feb 1997, Dean Gaudet wrote:
> > > It's too bad there's no way to pass a socket to another pid or we could
> > > have a cleanup process.
> > >
> > > Dean
> > 
> > There is! That's how NCSA 1.4 worked. In many OSes you can pass file
> > descriptors down a pipe to another process, using the I_SENDFD ioctl.
> > 
> Yes, it's not very fast, but it works. And we are talking about the tail
> end of error handling here...
> -- 
> chuck
> Chuck Murcko
> The Topsail Group, West Chester PA USA
> chuck@topsail.org
> 


Mime
View raw message