[NNTP] draft-ellermann-news-nntp-uri-02

[Charles Lindsey]

About 18 months ago, there were two competing drafts for updating the news
and nntp URI schemes:

         draft-ellermann-news-nntp-uri-00
         draft-lindsey-news-nntp-uri-00

The main difference, apart from style, was that mine advocated the use of
wildmats in place of the old '*' in the news scheme (meaning select all
groups) and the possibility of using a range as well as the number of a
single article in the nntp scheme.

There was also some discussion on this list which indicated that at least
the wildmats were a good idea. I suppose I should have followed up with a
revision of mine taking account of the discussion, but now Frank has
beaten me to it by publishing draft-ellermann-news-nntp-uri-02. He has, at
least, including wildmats (but not in their fullest form), which is
progress of a sort, but his text is IMHO over verbose and defines too much
details (e.g. precise characters allowed in a Message-ID - which is more a
matter to be sorted out by the NNTP server you connect to than by the
client making the request).

Anyway, what I would like from this list is some guidance as to what we
should be aiming for as regards the features to be provided. For details
on the text of the draft, the proper place to discuss it is on the
uri at w3.org list.

So, quick summary. The difference between the news and nntp schemes is
that the nntp is intended to work with article numbers, and hence
specifying the server is mandatory. So you can say

	news:12345 at example.com
or	news://nntp.server.example/12345@example.com
		to retrieve a single article by its message-id
	news:news.software.nntp
		to give access to all articles in that group
	news:*.test       {that is a wildmat}
		to give access to all the test groups
	news:
		to give access to all available groups (but it might come
		back with a full list of all 15,000 groups available :-(
plus versions of the above with a specific server such as nntp.server.example

	nntp://nntp.server.example/news.software.nntp/4567
		to retrieve a single article
+maybe	nntp://nntp.server.example/news.software.nntp/4567-4583
 or	nntp://nntp.server.example/news.software.nntp/4567-
		to give access to all articles in the given range (the
		last one could be useful to get any new articles, knowing
		that 4566 was the last one you retrieved on some previous
		occasion.

I think I take the view that any URI scheme which is intended to access an
NNTP server (as both these are) should enable access to as many as
possible of the facilities provided by RFC 3977 as can reasonably be
obtained from a single URI.

So retrieval by message-id maps into a single ARTICLE command.

Retrieval of what is available in a group or collection of groups maps
immediately into a LIST ACTIVE command (with a single group, a wildmat, or
no parameter at all). So the client interpreting the IRI just constructs
that LIST ACTIVE and deals with the list that comes back (letting the NNTP
server worry about whether the precise syntax is correct).

Retrieval by article number with the nntp URI would map into a GROUP
command followed by an ARTICLE command.

Retrieval of a range of articles in a group would map into a LISTGROUP
command, which now allows a range parameter (it didn't before).

As for wildmats, current URI implementations accept a variety of subsets
of them. But I think we can assume that implementations will steadily
converge on the new standard. Frank wants to limit them to
<wildmat-pattern> rather than <wildmat>, so he would not allow
	news.software.readers,news.software.nn
or	news.software.*,!news.software.nntp
but I don't see why. Those will be valid wildmats and will work as soon as
implementations catch up with the new standard. Likewise for <range>s.

It is also arguable that one might allow wildmats in the nntp URI (but
only if no <range> was present), just to enable people who intended to
work with article numbers to use the nntp URI as a one-stop-shop.

Anyway, opinions, please, and please join the uri at w3.org list if you want
to get involved more.	

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl at clerew.man.ac.uk      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5