Comment # 25 on bug 881506 from Petr Gajdos

(In reply to Jan Matejek from comment #24)
> Because: when you unload the module, apache will completely break because it
> won't be able to parse the "Require" statements. So if you want to unload
> it, you need to rewrite the config, and if you're rewriting the config, you
> can as well remove all "Require" statements and keep the module loaded.
> (unless there is something i'm missing, or unless we're potentially
> optimizing for server memory footprint, or something)

The user can use completely custom httpd.conf and system conf in old syntax.
Then he could consider to unload authz_*. Not sure if it is real-life example
:).

> Also please note that mod_authz_core is just the *core* module for the new
> access control method, and all mod_authz_* depend on it -- so if authz_core
> is unloadable, all the authz_* must be unloadable as well.

They all are as far as I can see.

> > For other configuration files:
> > 1. new syntax would work always with our default conf,
> 
> if you're talking about modules separate from the apache package, then no --
> new syntax is accepted but doesn't enable access if mod_access_compat is
> present.
> Modules must check for both and use both syntax variants.
> 
> > 2. old syntax should start to work after access_compat is loaded
> 
> again, no -- old syntax, without new syntax, is not sufficient to enable
> access, if new syntax disabled it
> 
> > 3. mix of new and old syntax would work as long as access_compat is loaded
> > and there is not contradiction (one permits and second forbids). Both types
> > of conditionals (comment 7, comment 18) should work.
> 
> comment 7 never works, because it is if/else. (enabling access for
> authz_core only when access_compat is unloaded; when both are loaded, this
> is wrong)
> to reliably enable both, we need if+if
> comment 18 always works (it's the if+if variant)

Sigh, you are right in all points (managed forgot from last time). Understood,
giving up :).