On Fri, 07 May 2010 15:51 -0400, "Joel J. Adamson"
<adamsonj@xxxxxxxxxxxxx> wrote:
> Joel J. Adamson <adamsonj@xxxxxxxxxxxxx> wrote:
>
> > William Pursell <bill.pursell@xxxxxxxxx> wrote:
> >
> > > Joel J. Adamson wrote:
> > > > Ralf Wildenhues <Ralf.Wildenhues@xxxxxx> wrote:
> > >
> > > >> If you don't need to work with Autoconf < 2.62, I suggest you add an
> > > >> AC_PREREQ([2.62])
> > > >
> > > > I had it, and then I discovered that the Beowulf cluster (emerald) at
> > > > the university here has version 2.59.
. . .
> 2) If I'm developing on the cluster with Ancient Autoconf and I need to
> fix *other* portability issues (related to the package maintainer's
> mysterious inability to update packages)
. . .
Concerning that "mysterious inability": there is a prejudice in the
sciences for replicable results, which software's potential
amplification of single bit differences can subvert. There is value in
being able to say "We ran this code and data under Scientific Linux 5.4,
with all public patches applied as of March 30, 2010". (Of course I've
no idea what's running on this Beowulf; I just pulled that out of the
aether.) Best would be to be able to hand someone the virtual machine
image for your environment, but that ideal can be out of reach,
especially on academic budgets.
All of which is simply to support Mr. Adamson's request as important,
serious, and difficult to avoid.
Best wishes,
- Stephen P. Schaefer
_______________________________________________
Autoconf mailing list
Autoconf@xxxxxxx
http://lists.gnu.org/mailman/listinfo/autoconf
