Andrew Gierth wrote about
<http://www.ietf.org/internet-drafts/draft-kohn-news-article-00.txt>
saying:
> it also fails to address any of the issues which have resulted in
> actual use of Usenet diverging from the previous specification.
I believe actual use of Usenet diverged from RFC 1036 because the latter
didn't support internationalized (i18n) headers or i18n newsgroup names.
As a result, most international users started sending a hodgepodge of
different, unlabeled charsets, which works within communities but scales
horribly for Internet-wide use. The obvious solutions are to either get
everyone to switch to UTF-8 or to get everyone to switch to
2047/punycode. There are pluses and minuses to both approaches. But if
there are other issues that I'm not aware of, please enlighten me.
> It also proposes changes (such as punycode group names) which have
> known compatibility issues with the Usenet backbone infrastructure.
I suspect you're talking about the fact that wildmat usage in NNTP (as
implemented today) will not do the right thing with file globbing within
component names (e.g., rec.arts.p?e* will not match to
rec.arts.xn--poe-punycodetexthere). However, I believe current wildmat
implementations will also often fail with UTF-8 encoded newsgroups,
since the ? will match only the first byte (out of up to 6) of a UTF-8
character. Further, I believe wildmat will do the right thing between
component names (e.g., rec.*.test) whether UTF-8 or punycode is used.
Therefore, whether UTF-8 or punycode is selected for newsgroup names, I
suspect the solution will be for the nntp-ext group to support an
optional enhancement to wildmat to support comparisons as a series of
Unicode characters, not as UTF-8 or punycode encodings.
In all cases, I believe punycode will break less things, and work across
more news and mail user agents, servers, and gateways than raw UTF-8
will. Only servers wanting to offer better searching of i18n newsgroups
will need to support the wildmat extensions, versus everyone needing to
upgrade to reliably handle raw UTF-8 newsgroup names.
If there are other issues that you are referring to, please let me know.
- dan
--
Dan Kohn <mailto:[EMAIL PROTECTED]>
<http://www.dankohn.com/> <tel:+1-650-327-2600>
