httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <>
Subject Re: [PATCH] must flush for CRLFs after POSTs
Date Tue, 28 Jan 1997 19:34:27 GMT
I'm going to resubmit this patch because I realised that btestread() is
redundant now that B_SAFEREAD exists.  I'll clean up what I did in buff.c
a bit as well.  I still think this is the easiest method to deal with

I think we can do better than Nagle, not to mention that it's generally
worthwhile reducing calls into the kernel.  What we're running into is
essentially the equivalent of calling fflush(stdout) before doing an
fgets() if you're waiting for user input. 


On Tue, 28 Jan 1997, David Robinson wrote:

> On Tue, 28 Jan 1997 wrote:
> > > But I'm not sure I like that solution. For one thing, mod_cgi is
> > > definitely the wrong place to put it, since POSTs can come
> > > anywhere. It would have to be done as part of the http_main loop that
> > > controls keepalives. And there's little difference between putting it
> > > there and in read_request(), as Dean as suggested.
> > 
> > Right.  I, for one, would have to clone whatever trick is added to mod_cgi
> > for this.  A generic solution in http_main_loop would definitely be preferable.
> Sure, I suppose any module can read POST data, therefore the fix for the
> extra CRLF data should be in http_main.c or wherever.
> I am a bit puzzled as to the original problem. What we are trying to
> do here is avoid small write()'s and hence reduce the number of packets
> by delaying the delivery of data. Isn't that precisely what the Nagle
> algorithm was supposed to achive?
>  David.

View raw message