[Linux-ha-dev] 3.0 thoughts

[David Lee]

On Mon, 1 Sep 2008, Lars Marowsky-Bree wrote:

> On 2008-09-01T15:48:07, David Lee <t.d.lee at durham.ac.uk> wrote:
>
> > o  "get rid of ConfigureMe and other autoconf abuse": they are two
> >    separate things:
> >
> >    o  autoconf abuse": sure, let's clean up!
>
> Are you volunteering? ;-)

Hmmm... not voluteering to lead, necessarily.  (My spare-time, volunteer
work on heartbeat has to be squeezed in around full-time job, family,
church activities, music writing, etc.)

But that config functionality is one I've worked on substantially in the
past (with my Solaris/Linux portability work etc.) and which I'm happy to
continue to try to be involved with in the future.

>
> >    o  "ConfigureMe": is a very useful 'front-end' onto configure for
> >       end-users who have to build their own: e.g where there is not a
> >       maintained/supported package/port available.  'autoconf' wants to
> >       put things into "/usr/local/{bin,etc,...}", but many OSes (and some
> >       of Linux-HA's own packages) prefer things elsewhere.  So a vote to
> >       preserve "ConfigureMe" in some form.  I'm happy to assist with such
> >       preservation and maintenance.
>
> I have certain reservations to ConfigureMe. More detailed:
>
> It detects system defaults. That's, in my book, a task for configure.
>
> I doubt that configure is _required_ to put things under /usr/local/bin
> instead of the relevant defaults as used by the system; if the user
> wants something else, they can still override them, afterall.
>
> This aspect of ConfigureMe is one which I consider quite wrong, and a
> pointless wrapper around configure and duplication of code.
>
> If this is stripped though, ConfigureMe essentially disappears and ends
> up being folded into "configure".

OK.  It's not that I necessarily want to preserve ConfigureMe "as is".
But it seems to contain useful functionality that would be worth
preserving somehow, somewhere.

On the Linux side, you (individually and collectively) do a great job in
building and maintaining downloadable packages for end-users.  And such
end-users, of course, don't need to get anywhere near any of this, not
even "configure".  Your "package-building" part of that process is to use
"configure" and/or "ConfigureMe" in a particular ways in your companies.

But there are several circumstances in which it is useful to be able to
say "if you are in environment-X, configure it like this; if in
environment-Y, configure it like that" etc.  For instance people (e.g.
Linux) who want to help debugging/development.  Or in our semi-supported
environments (e.g. Solaris) where there are not yet downloadable packages
from anywhere, but where we know that a simple "ConfigureMe" command (or
something equivalent in function and equivalent in typing simplicity) has
a much higher chance of working than having to remember a list of "--foo"
options.

So I suppose what I'm suggesting is:
 1. "ConfigureMe" contains valuable functionality for a significant
    minority of people;
 2. "ConfigureMe" has a valuable "type this simple, unadulterated command"
    component;
 3. Can we somehow preserve both those principles in any reworking of (and
    code redistribution between) "ConfigureMe" and "configure.in".

How does that sound?


-- 

:  David Lee                                I.T. Service          :
:  Senior Systems Programmer                Computer Centre       :
:  UNIX Team Leader                         Durham University     :
:                                           South Road            :
:  http://www.dur.ac.uk/t.d.lee/            Durham DH1 3LE        :
:  Phone: +44 191 334 2752                  U.K.                  :