httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: [PATCH] revised, final filtering patch. (ready for review)
Date Sun, 02 Jul 2000 04:02:52 GMT

On Sat, 1 Jul 2000, Greg Stein wrote:

> This is a revision of my patch to incorporate feedback from Roy, Ryan, and
> Dirk.
> Specifically, Dirk has an outstanding veto caused by the "char* handler"
> from my previous posted patch. This has been removed in this patch, so I
> would hope that Dirk likewise removes his veto.
> Other changes in this patch:

My concerns:

The buckets aren't complete enough.  This doesn't mean that we need to
implement more buckets.  It means that the bucket infrastructure is not
complete enough.  The buckets as currently implemented don't keep any
information about how they were created.  This means that filters later
down the chain don't know anything about the data.

How does a filter do anything?  All this implements is const char
*'s.  I also don't see anyway to actually chain buckets together.  The
buckets have a next and prev pointer, but nothing ever sets it.  Does this
mean its up to the filter to set these?

In general, I don't think this is it.  This is the previous patch except
it doesn't have the char * functions.  buckets may as well be char *'s in
this patch.  How are bucket's freed once a filter is done with them?  I
can't see how the data can be flushed to the disk and then removed from

At the very least, the buckets need to be allocated with
malloc, and they need a free pointer.  This design is just not complete.

The buckets are also likely to get VERY complex.  Having all of the bucket
types in one structure is going to get complex as soon as we start to
implement more of these things.

I really think we need Roy's bucket brigades.  This design just doesn't do
everything.  I'm not veto'ing, I'm just giving my opinion that this patch
isn't enough.


Ryan Bloom               
406 29th St.
San Francisco, CA 94131

View raw message