From: Russ Allbery (rra@stanford.edu)
Date: Mon Jul 02 2001 - 23:30:43 CDT
Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
> However, I still do not follow your argument. I would have thought that
> where there was the possibility to forbid some usage about which there
> was some doubt, it was always safest to forbid it, with the possibility
> of introducing it later (when the issues or requirements for it were
> better understood). Whereas it is _much_ harder to forbid a usage later
> on when maybe people have been actually using it for real.
Normally, I'd agree, but this case is special because it involves
canonicalization processes and a shared resource that has to be findable
by anyone interested in the topic. If we ban anything not in NFKC,
we're giving an implementation permission to canonicalize to NFKC. If we
then later change that specification, there will be a bunch of deployed
readers that can't deal with anything not in NFKC in newsgroup names, and
given the way Usenet newsgroup creation works, this will be sufficient
reason to not create such newsgroups (creating a self-perpetuating circle
that can persist for essentially forever).
Look at how hard it was to lift the 14-character limit in major
hierarchies, which *wasn't* actually coded in widely deployed software.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>