httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bill Stoddard" <>
Subject Re: FYI... working on reimplimenting the content_lenght_filter
Date Thu, 05 Jul 2001 20:07:10 GMT
If I create a CGI script that generates a lot of bytes, the cl-filter will read those
bytes into heap buckets and keep reading until the CGI is done sending before passing
anything down the filter chain. If the CGI generates 100MB of bytes, the cl-filter will
buffer it all before sending any of it down the stack. That's broken :-)  Well, the
nonblocking read stuff I just checked in earlier today -may- cause some bytes to get sent
down the stack, but that is not by design.

The modifications I am working on will do two things... make the cl-filter honor flush &
put a limit on the number of bytes the filter buffers before writing to the network.


> How is it broken?  Just saying it is broken doesn't give enough
> information.  It is very possible that the code isn't broken, it is just
> doing more than the name implies.
> Ryan
> On Thu, 5 Jul 2001, Bill Stoddard wrote:
> > Working on the nonblocking pipe read I realised that the content_length_filter is
> > quite broken.  I hope to get a patch out this afternoon.
> >
> > Bill
> >
> >
> >
> _____________________________________________________________________________
> Ryan Bloom              
> Covalent Technologies
> -----------------------------------------------------------------------------

View raw message