Re: Declaring a pre_config hook in a DSO will cause Apache to segfault

[Bill Stoddard <st...@raleigh.ibm.com>]

>
> Dean,
>
> I know you wrote it, but I have been digging in this code recently, and I
> think the pre_config hook was intended for MPM's only.  Because MPM's
> cannot be loaded dynamically

I hope we can change this. There are very good reasons to allow MPMs to be
dynamically loadable.

Bill

Re: Declaring a pre_config hook in a DSO will cause Apache to segfault

[Dean Gaudet <de...@arctic.org>]

my concern with dynamically loadable mpms is that the mpm provides the
threading model and threading models typically need inline functions and
other such optimiations which are compile time selected.  (there's
something lacking in the abstraction still, grr)

Dean

On Mon, 14 Feb 2000, Bill Stoddard wrote:

> 
> >
> > Dean,
> >
> > I know you wrote it, but I have been digging in this code recently, and I
> > think the pre_config hook was intended for MPM's only.  Because MPM's
> > cannot be loaded dynamically
> 
> I hope we can change this. There are very good reasons to allow MPMs to be
> dynamically loadable.
> 
> Bill
> 
> 

Re: Declaring a pre_config hook in a DSO will cause Apache to segfault

[Bill Stoddard <st...@raleigh.ibm.com>]

> > I hope we can change this. There are very good reasons to allow MPMs to
be
> > dynamically loadable.
>
> Such as?
>
The compelling reason for me is to enable shipping binary releases with the
option of using either the pre-fork MPM or one of the threaded MPMs. Apache
1.3 (and implicitly the Apache 2.0 pre-fork MPM) provides 90% of what most
people need and offers reliability that cannot be attained in a threaded
MPM. If a site is heavy into CGIs or mod_perl or their traffic doesn't
require the extra scalability offered by a threaded MPM, then they should
stick with pre-fork. I don't see that it would be difficult to make MPMs
dynamically loadable (handling the pre-config hooks, or ditching them
entirely appears to be the biggest challenge).

BTW, I agree this is not high on the list of things to get into Apache 2.0.
I wouldn't mind seeing it in a subsequent release though.

Bill

Re: Declaring a pre_config hook in a DSO will cause Apache to segfault

[rb...@apache.org]

> I hope we can change this. There are very good reasons to allow MPMs to be
> dynamically loadable.

Such as?

There are very few that I can come up with.  The only reason I can think
of to dynamically load MPM's, is so that one ISP can have difference
clients using different MPM's on different sites, without needing to have
the apache binary compiled multiple times.  While this is a useful thing
to have, I don't think it is an absolute necessity for the first release
of 2.0.

What other reasons am I missing?

If MPM's are dynamically laodable, then we need to either fix the
pre-config hook, or get rid of it altogether.

Ryan


Come to the first official Apache Software Foundation
Conference!!!   <http://ApacheCon.Com/>

_______________________________________________________________________________
Ryan Bloom                        	rbb@ntrnet.net
2121 Stonehenge Dr. Apt #3
Raleigh, NC 27615		Ryan Bloom -- thinker, adventurer, artist,
				     writer, but mostly, friend.
-------------------------------------------------------------------------------