On Fri, 28 Sep 2007, Max Bowsher wrote:
> Dan Christian wrote:
> > On 9/28/07, Ben Collins-Sussman wrote:
> >> To be utterly clear: from the user's point of view, there is no
> >> compile-time or run-time dependency on any particular version of swig.
> >> The svn release tarball ships with bindings that are
> >> ready-to-compile, with no external dependencies.
> >>
> >> So it's irrelevant what version of swig happens to ship in any
> >> particular linux distro. The *only* people who need a certain version
> >> of swig available are svn developers building from a working copy.
> >
> > It was always clear that SWIG was optional for users and developers
> > who don't touch the bindings.
> >
> > It's just that many developers use the same slow release distros as
> > the users (its called a "day job" :-). So touching the swig bindings
> > (with a 1.3.31 requirement) means building and installing swig from
> > scratch.
>
> ./configure
> make
> make install
>
> done.
Max, Dan was referring to actually *changing* the SWIG bindings, which
requires SWIG and autogen.sh. He may be doing this as part of $day_job,
and tied to a specific platform where a more recent version of SWIG
isn't readily available, and must be custom-built.
> > Which is fine if you get something for it. So far, no one has said
> > what is the gain is of requiring 1.3.31 instead of 1.3.27. There were
> > comments that 1.3.25 was a bit change, but nothing more specific than
> > that.
>
> * Discarding compatibility code.
> * Developers not having to think about differences across a broad range
> of versions.
> * Developers not having to *TEST* so many versions.
>
> > I'm just trying to keep the developer barrier to entry as low as is
> > practical ('cause it might make my life easier :-).
>
> Weigh the possible need for a new developer to download swig and run
> "./configure && make && make install" against the expectation of
> producing patches which are functional across a broad range of swig
> versions - I think it's justifiable to say that tightening up the
> allowed version range actually lowers the real developer barrier a bit.
Me too, though perhaps not going so far as dropping all but the most
recent version.
--
Daniel Rall
- application/pgp-signature attachment: stored
Received on Sat Sep 29 03:21:10 2007