httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@rowe-clan.net>
Subject Re: cvs commit: apr/network_io/unix sendrecv.c
Date Mon, 26 Feb 2001 22:22:52 GMT
From: <rbb@covalent.net>
Sent: Monday, February 26, 2001 8:53 AM


> On Mon, 26 Feb 2001, Greg Stein wrote:
> > I don't understand this checkin at all...
> >
> > mod_core.h was to be private to the mod_core module. Now it is public, yet
> > located in a different directory. If it is to be public, then it should just
> > go into httpd-2.0/include. Hell, why even make it exist, given that we have
> > things like http_protocol.h?
> >
> 
> mod_core will become private again in time.  Right now, things are in
> flux, and to keep things building I had to make it public.
> 
> > Second, I see that server/protocol.c was created and a ton of stuff pulled
> > from modules/http/http_protocol.c and shoved into protocol.c. However, most
> > of those functions *are* http-specific. So why'd they get moved? If the
> > intent was to keep http-specific stuff in modules/http, then what's up?
> 
> Any function I moved is a function I don't believe is protocol specific at
> all.  They are functions that are part of how Apache works, and to put
> them in the http module means that no other protocol module can use them.
> Having said that, I am perfectly willing to be wrong.  Things will likely
> move back and forth until we get this right.
> 
> BTW, my goal is to compile an Apache without mod_http, but with another
> protocol module, and still have it work cleanly.  Currently, that is not
> possible.

Ryan...

if you are hanging out mod_core.h for a few days - could it _please_ move to
httpd-2.0/include till you finish and it gets its privacy back?

Second... 

can we do _something_ about the symbol switching you have going on?  I see
a bunch of missing AP_DECLARE()'s - or a few too many.  I can't grok if these
fns will become private and restricted to mod_core, and never move around,
or if they will be accessable.  If they are supposed to be accessable, even
to 'intimate' modules like loadable protocols, please put the AP_DECLARE()s
back in ... otoh if they will not be exported please remove the AP_DECLARE()s
from the header.

Thanks, we will bear with you as long as we know where we are headed :-)

Bill


Mime
View raw message