splitting courier-authlib into master+slave ports

[Oliver Lehmann]

Hi Jose,

Jose M Rodriguez wrote:

> I don't think so.  A common user expects couier-authlib be capable of 
> reading system passwd 'out of the box'.  The rest must be selectable.

What is a "common user" and how can you know what that user expects?

> Well, can we route the PRs to you? I Expect 'some mail' if 
> courier-authlib can't read system passwd 'out of the box'.

As I am the maintainer, you can route them to me, no problem. I'll
redirect them to UPDATING where I'll describe that there is no auth.
method installed by default and the end-user has to choose a sub-port
which fits his needs.
When courier-authlib gets split into base and auth. ports this should be
done as clean as possible which means not letting some auth methods in the
base-port and some not. That would not fit my definition of "clean". I
see the point that it may cause confusion "I Installed sqwebmail and it
doesn't work, I can't login!" But I don't see a way to prevent that when
the seperation should be clean. If I'm getting to much user's feedback
that this is to confusing, I might reconsider taking Yarema's
Makefile.opt idea.  Include that in every port that depends on courier-
authlib and choose there what auth. methods to install.

But I would say it should just be done now and we'll see what happens. Or
we can discuss wnow hat happens w/o getting it done for ages... (imho)

-- 
 Oliver Lehmann
  http://www.pofo.de/
  http://wishlist.ans-netz.de/