>>>>> "Russ" == Russ Allbery <[EMAIL PROTECTED]> writes:
Russ> Akim Demaille <[EMAIL PROTECTED]> writes:
>> I'm sorry, but I don't quite agree: your picture lacks the Autoconf
>> macro archive. Its existence makes a huge difference with the
>> former state of Autoconf.
Russ> A macro archive by itself is useful but not sufficient to
Russ> replace a set of useful macros distributed with autoconf. The
Russ> advantage of having the macros distributed with autoconf is that
Russ> one can then know for certain precisely which macros are
Russ> available in a given version and not include a whole bunch of
Russ> general macros in every package.
I don't think we statically disagree, we just differ on the dynamics
:)
Russ> It may be worth considering whether autoconf should no longer
Russ> serve as the general-purpose macro repository as well as the
Russ> infrastructure, and instead a separate distribution with
Russ> specific version numbers be created that will.
At least technically this is what is currently happening. Was used to
be reserved to the Autoconf maintainer is now available to end users.
OTOH, it is also good that Autoconf keeps in touch with the real
world, and swallows some specific macros, let it be to `dictate' the
good coding style and the proper use of the API.
Russ> It may be worthwhile considering whether to just break backwards
Russ> compatibility to some degree and fix things once and for all,
Russ> resulting in an autoconf 3 release.
I sentimentally 100% agree with you, but unfortunately I sort of
understood the incredible momentum users of Autoconf have. At first,
since they are maintainers, I thought they'd easily agree on deep
changes in the interface of Autoconf, but I had to realize this is not
the case[1]. So now I'm more in favor of a soft transition (which
does not mean there are no earth quakes underneath :).
Akim
[1] I am criticizing nobody. I'm amongst the most fervent detractors
of Autoconf because it requires too much knowledge from the
maintainers who use it, so I am certainly not criticizing maintainers
who don't want to learn again something they just shouldn't know. It
just happens to be different from what I first imagined.
