Re: choice of implementation language

[Ralf Wildenhues]

Hello,

> Jim Meyering <address@hidden> writes:
> > So I conclude that the choices are
> >
> >   Perl
> >   Python
> >   Ruby

FWIW, my preferences are: sticking with what we currently have, or Perl.
Python is installed on less than half the systems I test on, and Ruby on
virtually none other than the Linux ones.  (Transferring autotools
output to them would be mean a development slowdown.)

I'm not a gnulib maintainer, but I don't see gnulib-tool maintenance as
being such a huge issue.  Certainly my patches touched code virtually
unchanged for several years.  I do take objection to being called on
premature optimization, after having done lots of measurements to
delimit the hot spots in a bootstrapping procedure, and with the code
not looking like it was in its early development stages.  I maintain
that my not-applied patches can be redone in a fashion that is
maintainable and (just as) safe and fast (e.g., by combining all the
added eval into just one; the longer sed script is hardly one that will
need much maintenance, as its job is rather straightforward; nothing
that can't be tested efficiently with a few test suite additions).

Cheers,
Ralf