Return-Path: Delivered-To: new-httpd-archive@hyperreal.org Received: (qmail 12740 invoked by uid 6000); 3 May 1999 12:58:32 -0000 Received: (qmail 12730 invoked from network); 3 May 1999 12:58:31 -0000 Received: from zap.ne.mediaone.net (24.128.120.74) by taz.hyperreal.org with SMTP; 3 May 1999 12:58:31 -0000 Received: (qmail 3582 invoked by uid 1000); 3 May 1999 12:58:28 -0000 To: new-httpd@apache.org Subject: Re: Module initialization and per-directory config directives References: From: Ben Hyde Date: 03 May 1999 08:58:27 -0400 In-Reply-To: Zeev Suraski's message of "Sat, 1 May 1999 23:31:42 +0300 (IDT)" Message-ID: <87k8uqjong.fsf@zap.ne.mediaone.net> Lines: 32 X-Mailer: Gnus v5.5/XEmacs 20.4 - "Emerald" Sender: new-httpd-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org Zeev Suraski writes: ... > directives were being parsed before module startups ... > ... the callbacks so that they always call PHP's startup function > before going on. ... It is possible if you proposed a patch it would get in. I think every significant module I've got or read does just that. Every configuration directive calls out to the module begin init, and then latter when Apache calls the module so called module init, it does he "end init". There are two reasons why there is not a "begin init" call. First it is a slightly delicate to design since you have to decide what to do in all four (or more?) senarios. 1. Directive reading during requests. 2. Directive reading phase 1. 3. Directive reading phase 2. 4. Restart directive reading. Second there developed during the last release a reluctance to adding new callbacks. The design for more hooks built up, encryption export issues made the water muddy, API stablity advocates spoke up, tempers flaired. Ralf implemented EAPI to get them for his needs. I did the same, I suspect others did too. - ben