Re: sfio support

* Ricardo Signes <perl.p5p@rjbs.manxome.org> [2011-12-21T10:40:44]
> * Phong Vo <kpv@research.att.com> [2011-12-20T10:46:27]
> > I am interested in learning what problems you have with the library.
> > If I can help, please let me know.
> 
> Great!
> 
> I'm not the guy to give you those, but I will talk to a few of the people who
> can and either gather information or urge them to reply.  Thanks for the
> email.

Hi, it's me again.  Thanks again for your offer.  Things boil down to this:

Right now, perl can be built with either PerlIO (its own IO system), stdio, or
sfio.  If anything but PerlIO is chosen, one's perl is massively limited in its
IO features to the way things were in 2002, more or less.  It is in this
context that our sfio has fallen into failure to build.

So, while it would be nice if it still built, it seems like all it would do is
get us into a state where there'd be one more working option to get you to a
place that you really don't want to be.  I think it would be a poor use of
time.  Instead, we're going to work on improving PerlIO and stop working about
not having PerlIO.

We did talk some about whether sfio could be used to outsource some of the
things that PerlIO does, but it looks like we'd be stuck on a few key points.
Most importantly, we would still need another implementation for anything for
which we used sfio, because sfio is not license compatible and could not be
bundled.  (I am also doubtful that it is as portable as we require.)

So, it seems inevitable that the current sort of support for sfio in perl is
going to be deprecated -- no one would want to use it as is, even if it worked.
We may revisit sfio in perl in the future, though.

-- 
rjbs

• sfio support by Phong Vo
• Re: sfio support by Leon Timmermans
• Re: sfio support by Steve Peters
• Re: sfio support by Ricardo Signes
• Re: sfio support by Ricardo Signes