httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rodent of Unusual Size <Ken.C...@Golux.Com>
Subject Re: Options inheritance
Date Wed, 28 Jan 1998 16:26:19 GMT
Randy Terbush wrote:
> There are some other config parameters that don't get propigated
> as I remember. I'm beginning to thing that perhaps we need
> an explicit directive that would tell the server to inherit or
> not inherit the current config. Or perhaps a 3rd (4th, 5th) parameter
> that would export a setting globally?

I've been toying with the idea of adding yet another slot to the
module structure: a callback to allow the module to return some
sort of configuration status.  Perhaps something like:

 char *req_config(request_req *r);

It would be up to each module to provide such a hook if desired,
and the return value would be free-format text describing the
module's configuration for that request.

This has synchronisation problems (such as when it can be called),
so maybe a better method is a couple of reserved slots in r->notes,
such as "<modulename>_r" and "<modulename>_s".  The server and
directory config/merge routines could set these appropriately.

Once again there's the issue of who could use these and when; maybe
a special content-handler module that looked these up and
displayed the results?

I'm just writing off the top of my head, but I think the second
approach is better - the missing piece is setting the convention of
the keys in the r->notes table for this information.  Retrofitting
the standard modules would be relatively (!) simple; the query
module likewise.

Just winging it.. because I've often wanted to be able to verify
settings with something like "http://host/analyse?/some/location/foo"
and see what the server thinks the overrides and options are for that

#ken	P-)}

View raw message