httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bill Stoddard" <stodd...@raleigh.ibm.com>
Subject Re: AP_FTYPE_CONNECTION == BUFF + IOLs
Date Fri, 25 Aug 2000 20:33:42 GMT

> On Fri, 25 Aug 2000, Rodent of Unusual Size wrote:
>
> > rbb@covalent.net wrote:
> > >
> > > This question had very little to do with the removal of IOLs.
> >
> > It had EVERYTHING to do with it.  You took them out, and broke
> > Windows utterly because you didn't supply sufficient replacement
> > functionality at the time of the removal.  You should have done
> > your pipe bucket *first*.
>
> No.  That complaint was in a completely different message.  I can
accept
> that complaint.  That was actually a veto, that has since been
removed.
>
> The message that started this thread had nothing to do with the IOL
> removal.  It had to do with why we are trying to use filters to
repalce
> IOLs and BUFFs.  While the two are related, they are not the same.
>
Sorry I didn't explain sufficiently... I vetoed the patch because it put
the server in a completely hosed up state for no reason. I would not
have vetoed a patch that removed IOLs if the pipe bucket, et. al. was in
place to replace the function that was removed. You removed a critical
piece of code and replaced it with what you admit was a hack and that
hack did not work on all platforms. I then questioned the need to remove
IOLs at all. You hack sort of led me to that conclusion. You replace
ap_bpush_iol(), which can handle pipes, files or sockets, with
ap_push_socket(). Then you proceed to push a pipe onto an apr_socket_t.
What did you accomplish by removing IOLs and replacing them with a hack
(other then break Windows :-)?

I am really interested to see how mod_cgi is implemented using filters
and BBs. In the meantime, I am taking a break from Apache development on
Windows for a while ....

Bill


Mime
View raw message