apache2, apache20: mpm variant issues

[ryandesign (Ryan Carsten Schmidt)]

I've never looked into the mpm variants in apache2 (and apache20) before, but now that I do, I see a number of issues.

1. The three mpm variants are mutually exclusive so they should be marked as conflicting with one another
2. They should have variant descriptions
3. One of them should be marked as a default variant if no other mpm has already been chosen:
   1. Either the preforkmpm variant should be default, since the comment in the portfile reads "This ends up being the default mpm if no other is chosen"
   2. Or the workermpm variant should be default, since the comment for that variant reads "This mpm is probably preferred (maybe we should make it a default variant?)"

It may be wise to do 3.1 now and increase the revision, then wait a few months to ensure most users have upgraded to that version, and only then, if 3.2 is the actual desired end result, do 3.2 then. That way, existing users who already have the preforkmpm will not suddenly be switched to the workermpm, which might have adverse effects (e.g. don't you have to recompile php and all your php modules if you change apache mpms?)

[nerdling (Jeremy Lavergne)]

PHP has ​a juicy warning that if you have to use MPM, use prefork; if you need threads, use fastcgi.

[nox@…]

We should not change the default MPM, Apache has chosen the prefork one on UNIX, let's do the same.

[ryandesign (Ryan Carsten Schmidt)]

Fixed apache2 in r56514 (maintainer timeout) using a reworked version of Anthony's patch from #20743.

[ryandesign (Ryan Carsten Schmidt)]

Merged to apache20 in r56515.