httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From efr...@ncsa.uiuc.edu (Beth Frank)
Subject Re: Server push problem
Date Wed, 25 Oct 1995 19:39:11 GMT
> 
>  
> > >What problems would we run into if we stop buffering CGI output?
> > 
> > A lack of any future enhancement? Buffering is required for
> > * protection of the connection file descriptor from the script
> > * implementation of transfer-encodings for persistent connections
> > * HTTP-NG perhaps?
> > 
> > However, that the server could flush its buffers when no data has been read
> > from the script for a while (5 seconds, say.)
> 
> An off the top of my head idea...
> 
> Can we let scripts send back a "for apache only" header which asks
> that the buffer size be reduced?
> 
> I've got some scripts which produce reasonable amounts of output but 
> when accessing them over a transatlantic link, it's painfull to watch...
> data jerks through in 4k blocks. Now if the script could ask to send 
> in say, 512byte blocks, the output would (in theory maybe) flow 
> 8 times more smoothly ;-).
> 
> If there's no technical reason to counter this suggestion, then it'll
> add more "user control"... something I very much like to encourage in 
> Apache.
> 
> 
> rob
> --
> http://nqcd.lanl.gov/~hartill/
> 

If you'd like see this kind of feature implemented more widely than
just in Apache, I suggest you mention it on the http-wg mailing list
(http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, I believe the subscription
address is http-wg-request@cuckoo.hpl.hp.com).

-- 
		Elizabeth(Beth) Frank
		NCSA Server Development Team
		efrank@ncsa.uiuc.edu

Mime
View raw message