ietf-nntp Newsgroup Name Length Question

[Martin J. Duerst]

On Wed, 15 Jan 1997, Stan Barber wrote:

> Ken writes:
> > Defining a group name as a non-empty sequence of characters not
> > including control characters, white space, wildmat wildcards, commas,
> > and whatever I've forgotten should be fine.
> 
> I like this. We'll need to define the "whatever I've forgotten" part, but
> this seems like a reasonable approach.

Here is something for "whatever I've forgotten": Internationalization.

There are good reasons why the widely distributed newsgroups have
ASCII-only names, but there is no need to restrict this for newsgroups
that are used more locally.

There are various ways with which non-ASCII usegroup names could be
allowed. One would be to use UTF-8, an ASCII-compatible encoding
of UNicode/ISO 10646, the universal character set. This would be
along the recommendations of the IAB charset workshop
(see draft-weider-iab-char-wrkshop-00.txt) and would work together
well with other areas such as ftp and URNs. If the syntactic limitation
to ASCII in newsgroup names turns out to be too hard to change, there
are other ways to introduce multilingual names. For examples, please
see draft-duerst-dns-i18n-00.txt or RFC2060, section 5.1.3.

What is particularly important is that whatever way of internationalization
is choosen, it can affect "reasonable" name length (in terms of bytes)
considerably. For example, while the limitation of DNS to 63 bytes per
part and 255 overall is extremely generous for English domain names
that can be represented directly in ASCII, it is considerably
less generous for characters beyond ASCII that have to be encoded
with more than one byte. A restriction to 14 bytes per part could
mean a restriction to only 3 characters or even less, depending on
the character's place in Unicode/ISO10646 and the choosen encoding.

So while it makes sense to give indications as to the length of
group names, these should not only be based on the list of existing
newsgroups.

Regards,	Martin.