ietf-nntp newsgroup production

[Russ Allbery]

Chin Chee-Kai <cheekai at SoftML.net> writes:

>>>>>> newsgroup-name  = component *( "." component )
>>>>>> component       = component-start
>>>>>>              *( component-start / component-other )
>>>>>> component-start = Un-lowercase / Un-digit
>>>>>> Un-lowercase    = <Unicode Letter, Lowercase> /
>>>>>>           <Unicode Letter, Other>
>>>>>> Un-digit        = <Unicode Number, Decimal Digit> /
>>>>>>           <Unicode Number, Other>
>>>>>> component-other = "+" / "-" / "_"

> Does "Other" include the character "-", and "Unicode Letter" presumably
> not include character "-"?  Does the "< xxx , yyy >" notation mean the
> character sets xxx followed by yyy?

No, and no.  <Unicode Letter, Other> is one set of characters, consisting
of all "other-class" Unicode letters (not uppercase or lowercase).
Similarly, <Unicode Letter, Lowercase> is all Unicode lowercase letters.

A component consists of component-start followed by any number of
occurences of either component-start or component-other.  component-start
gets all the letters and digits and component-other includes "-", "+", and
"_".  (There are some additional notes in the USEFOR description about
things like components consisting solely of digits.)

> If these are true, then you're right.  It'd help if you had included
> complete definitions for the <...>s.

It's an extension to RFC 2234 for things that are hard to write production
rules for.  Think of it as a "character class"; the text inside <> is an
English description of the characters making up the class.

> I don't quite follow here.  Implementors of servers use this draft/RFC
> to decide whether a newsgroup reference is "legal" or not.

NNTP is not required to determine the fine points of the syntactic
validity of newsgroups; it doesn't hurt anything to accept strings to a
GROUP command that aren't valid newsgroup names.  The server will just
never create those groups (and the rules on what groups can be created do
have to come from RFC 1036 and its successors; it's none of the business
of NNTP).  The only requirement for the NNTP specification of newsgroup
names is that it be no more strict than the RFC 1036 successor description
and that it not permit anything that's obviously wrong (like newsgroup
names containing whitespace).

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>