From: Russ Allbery (rra@stanford.edu)
Date: Wed Oct 20 1999 - 17:47:12 CDT
Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
> 1. Should not the FWS be CFWS? Currently, we only forbid comments in
> places where implementation speed might be critical, notably in
> Newsgroups and Path.
Backward compatibility problem. If you use:
Control: newgroup (create this) comp.foo
most existing servers will probably take that as a request to create a
newsgroup named "(create" with flags "this)". We can say that servers are
required to understand and ignore comments but aren't allowed to generate
them, but I don't think we should go farther than that at this time.
> 2. Should not a verb be a token (that seems the current fashion for
> keywords that have to be recognised by parsers)?
Sounds fine to me.
> 3. I am no longer so sure about the restriction to quoted-string. It
> would mean for example:
> Control: newgroup "comp.@*#@#*"
> where @*#@#* is some collection of UTF-8 characters for some chinese
> newsgroup. But we do not insist on quoting them in a Newsgroups line
> (the syntax of <newsgroup-name> is quite clear, and I would expect to
> re-use <newsgroup-name> when writing the syntax of the newgroup control
> command). OTOH, it would be nice to be able to do a preliminary parse on
> all Control commands by a single straightforward parsing process.
Reusing the syntax of the Newsgroups line makes sense to me; doesn't it
already disallow whitespace in newsgroups? If so, provided the rest of
the Control header is whitespace-delimited the way it is now, you can
still do a simple preliminary parse.
-- Russ Allbery (rra@stanford.edu) <URL:http://www.eyrie.org/~eagle/>