1.3 modification question

[hbeaumont hbeaumont <ah...@gmail.com>]

Hi,

Please point me to the proper list if this is not the right list for
1.3.x issues.

(Disclaimer: I know there are better ways to do what I'm about to describe)

Does anyone see a problem with the patch/hack/mod:

When a request comes in, the main server checks the destination
virtual domain and suexec defined user then spawns the child process
as that user (instead of apache/nobody/etc.).

All children are set to allow Keep Alive requests but must die after
all requests are served (so
that other virtual domains are not served with that as the wrong user name).

I know there is suexec and modules for 2.x to allow this sort of
thing. This is more just for my own learning.

Are there any obvious flaws? I would appreciate any advice if this is
just totally not possible.

I think it would be also be interesting to keep open children alive
for popular vhosts.

Again, just for own fun and research. I know that this type of thing
is not for inclusion into the httpd code and I know there are plenty
of other ways to do the same thing.

For those interested - I figure the extra resources spent launching
new httpds (since each child could only serve one request plus its
keep alives) wouldn't be any worse than the resouces used to span
fastcgi'd ruby/php/whatever.

Thanks for any comments or direction to a better place to ask.

Re: 1.3 modification question

[hbeaumont hbeaumont <ah...@gmail.com>]

On 8/22/06, Joshua Slive <jo...@slive.ca> wrote:
> On 8/22/06, hbeaumont hbeaumont <ah...@gmail.com> wrote:
>
> > Are there any obvious flaws?
>
> Horrible performance?

agreed.

>
> >
> > Thanks for any comments or direction to a better place to ask.
>
> Search the archives of this list and the users list for any thread
> mentioning "perchild".  You'll find the challenges thoroughly
> discussed.  This is all for 2.x, of course.  Everything would be much
> worse in 1.3 because fo the less-flexible and less-modular code base.
>
> Few people are interested in hacking on 1.3 anyway.  It hasn't had
> significant development in over four years.
>

Agreed. This is all just to answer some questions I had for myself and
a (perhaps pointless) little project for me.

Thanks for the input though!

Re: 1.3 modification question

[Joshua Slive <jo...@slive.ca>]

On 8/22/06, hbeaumont hbeaumont <ah...@gmail.com> wrote:

> Are there any obvious flaws?

Horrible performance?

>
> Thanks for any comments or direction to a better place to ask.

Search the archives of this list and the users list for any thread
mentioning "perchild".  You'll find the challenges thoroughly
discussed.  This is all for 2.x, of course.  Everything would be much
worse in 1.3 because fo the less-flexible and less-modular code base.

Few people are interested in hacking on 1.3 anyway.  It hasn't had
significant development in over four years.

Joshua.