Return-Path: Delivered-To: new-httpd-archive@hyperreal.org Received: (qmail 16865 invoked by uid 6000); 8 Sep 1999 11:30:09 -0000 Received: (qmail 16841 invoked from network); 8 Sep 1999 11:30:06 -0000 Received: from kurgan.lyra.org (root@208.192.43.9) by taz.hyperreal.org with SMTP; 8 Sep 1999 11:30:06 -0000 Received: from kurgan.lyra.org (IDENT:gstein@localhost [127.0.0.1]) by kurgan.lyra.org (8.9.2/8.9.2) with SMTP id EAA17922 for ; Wed, 8 Sep 1999 04:21:08 -0700 (PDT) Message-ID: <37D646A3.65AA1ED9@lyra.org> Date: Wed, 08 Sep 1999 04:21:07 -0700 From: Greg Stein X-Mailer: Mozilla 3.01 (X11; I; Linux 2.0.28 i586) MIME-Version: 1.0 To: new-httpd@apache.org Subject: Re: Apache using apr. References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: new-httpd-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org Dean Gaudet wrote: > > On Fri, 27 Aug 1999, Greg Stein wrote: > > > We should use contexts so that the user data can be carried around. Dean > > suggested added a void* to the pool structure. That is a bit silly as > > that means pools are no longer just memory handling types, but include > > other stuff. The best label to apply for that case is "context". > > note if you see my other post about supporting all types of allocation > within one pool you'll see that it'd be cool for the different allocators > to have a void * that they can allocate within the pool... I'm not sure that I follow what you're getting at here. Would the allocator use the void* internally? Well, sure... that makes sense. I would expect them to have some kind of private storage in there. Or do you mean something about exposing the void* to users? In that case, then you're monkeying up the abstraction between allocators and context storage. > and as far as usage for the void *, we need some sort of unique identifier > for each user of the pool. ken suggested the module number method, but of > course that doesn't work within APR, because modules are an apache thing. atoms. At a minimum, that damn "table" type really has to change. The "values are strings" thing is just a horrendous pain. I want a name => void* mapping. Hell, I want binary-blob => void* mappings. Where are those Python dictionaries when you need 'em? :-) Cheers, -g -- Greg Stein, http://www.lyra.org/