httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <gst...@lyra.org>
Subject Re: cvs commit: apache-2.0/src/main http_core.c http_protocol.c util_filter.c
Date Tue, 29 Aug 2000 23:06:39 GMT
On Tue, Aug 29, 2000 at 03:24:40PM -0700, rbb@covalent.net wrote:
> 
> Before people start yelling that this patch was vetoed, I know that.  I am
> committing anyway, because I have asked for two weeks for a technical
> reason for the veto, and I haven't seen one yet.  I got sick of waiting,
> so I decided to just commit.  This may be an unpopular decision, but I am
> hoping it gets things moving again, because we keep stagnating.

1) I gave what I thought was a valid technical reason, but you chose to not
agree. That is your perogative, but it is also mine to veto what I believe
is an improper change.

2) I disagree that you can simply say that a veto goes away after time. If
you wanted to push for a reason, then ask again. I've been busy (as
evidenced by my lack of emails here) and just haven't cared to deal with the
issue. But asking once, waiting a while, then unilaterally deciding a veto
is thereby closed is way wrong.

3) stagnating? Feh. Are you trying to say that your SSI filter can't be
written and used? Are you saying that Jeff's xlate filter can't be written
and used? Or how about the chunking filter that was added? The code for that
has had review and fixes applied and is now used by Apache. That is a lot of
progress. You have an issue with how these are inserted, but that is a small
part of the overall problem: in no way is filtering "stagnating".


"technical reason" you say. Fine. I'll state it again:

Your "LIFO change" (this stuff really isn't a LIFO because we only
add, there is no "out" ordering) is solving the problem you had with the
CORE and CHUNK filters, and shoves a similar (reverse) problem on the
non-connection filters. Namely: connection filters appear to be a bit easier
with "prepend"; content filters appear to be easier with an "append". We
chose one: append. Your patch doesn't solve the overall problems, and
introduces a discrepancy between the timing of insertions via
"insert-filter" and the process used by ap_add_filter (if you order your
hooks in a particular way, then the filters go in reversed).


> If I pissed people off with this commit, sorry.  I would ask that this
> commit not be backed-out until there is a valid technical reason for the
> veto, because I would like to move forward, and I couldn't go anywhere
> with the broken code that was there.

You certainly damn well could continue with the code that was there. It was
working quite fine. The "+ 1" in there was just a temporary measure until we
introduced the updated set of groups. As I mentioned, I haven't been pushing
on the full resolution there because the code was working fine, and I've
been doing some other stuff.

The code implemented the grouping of filters that we discussed at length
back in April to June, and at the pow-wow. Your checkin completely threw
that stuff out.

-g

-- 
Greg Stein, http://www.lyra.org/

Mime
View raw message