httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <>
Subject Re: I'm changing my mind about config modules
Date Fri, 14 Apr 2000 18:46:52 GMT
On Fri, 14 Apr 2000, Manoj Kasichainula wrote:
> One of those minor epiphanies that happens when you're not thinking
> about anything at 3:30AM. Just noticed Dean's message basically saying
> the same thing, BTW. Cool!
> Maybe it's because I'm not paying as much attention lately, but the
> configuration module scheme I've seen proposed just feels complicated.

The only thing complicated (and not even that) is that we broke it down
into several easy steps. I'll review:

Phase 1

1) build config tree during normal config processing. Done.
2) call modules based on tree rather than during config processing.
3) move config processing "out" of src/main/. processor calls back through
   an API to construct the config tree.

Phase 2

1) define XML schema for describing modules and directives
2) based on (2.1), add validation to the callbacks

Phase 3

1) Make config modules loadable and selectable, allow multiple inputs

Nothing here is complex. Phase 1 should be done for any number of reasons,
least of which includes modularization (I mean nice divisions between
code, not necessarily "Apache modules") and to allow for things like your
simplified syntax idea in the rest of this note.

Phase 2 is a little tougher, but only from a discussion point. Everybody
has their own ideas on what the XML schema should look like :-). But the
code isn't complicated. Providing XML schemas will give us at least a
couple things: better ways to describe features of modules, mechanisms for
external processes to learn about the installed modules and how to config
them (e.g. GUI config tools), and increased validation capabilities.

Phase 3 is the part that you and Dean seem to not like -- allowing other
config modules to be loaded and used. Correct?

> If we're not going to make dynamic configuration the primary goal,

Not in this round. Dynamic config has a huge number of thorny issues. By
creating the config tree, we provide a basis for adding dynamic config
later on. The several phases and steps above provide a basis for dynamic
config, which we felt was simply too difficult to tackle at this point.
The delta from "today" was too great.

If you want dynamic config, then by all means. But I would recommend that
we proceed on the current path to get you there.

>... more stuff about simplified config syntax ...

The rest of this is pretty moot. If the group feels we should move towards
a simplified syntax, then great. It will be quite easy once Ryan is done
with this group of changes.

> We should still do stuff listed in Ryan's original message; some of it
> would still be applicable under this scheme. Think of this as moving
> the API boundary from functions inside Apache to text files outside
> Apache. In fact, this was partially inspired by the sysctl/proc debate
> on linux-kernel.

Well, at least you agree here :-)


Greg Stein,

View raw message