> I'd suggest extending the syntax to read as follows
> news:comp.sys.foo comp.sys.foo on default system
> news://nntp.archive.msen.com/comp.sys.foo nntp.archive.msen.com
> news://nntp.archive.msen.com:1990/comp.sys.foo (on port 1990)
> with similar extensions for referencing individual articles.
> - Ed
The syntax would certianly fit in with the UDI format -- however, are the semantics
well defined? In general, the whole point of news is that it is held locally,
avoiding millions of WAN accesses.
If you put in a node name, then you are changing the way the protocol works
altogether. This might be conveient, but its not really "news". You're saying
that we can use NNTP as a file retrieval protocol. Obviously, once you have given
out a reference like that above you have to make nntp.archive.msen.com available
for everyone in the world.
Maybe this is a way to solve the news archive retrieval problem. It isn't done
that way at the moment of course: Its a big headache right now. Messages are tared
and compressed and put on some machine under the date of the message -- you have to
know which newsgroup the message was sent to, and then look the archive hostname up
in a list which doesn't (correct me?) exist. (Another use for the X-500 directory?)
It doesn't fit very well into the news model, all the same. For example, when you
find a reference to another newsgroup/article on your CD rom, There's no way of
knowing whether you should look it up on the CD rom or on a "live" news server.
We really need some NNTP extensions to insist that the message-id can carry some
hints as to where it might be archived -- unfortunately I missed the NNTP session
at the IETF but I know that Eliot Lear (ietf nntp wg chair) for example is thinking
about such problems, and indeed from talking to him I got the impression that the
NNTP group's discussions were overflowing into the retrieval and resource discovery
It's not _trivial_ to put it in the code for (One will have to keep a cache of
network connections to hosts) so I won't do it now. I'd point out you could set up
an HTTP server to serve the news data, although it would have (currently) to
convert the news format into HTML which it (currently) doesn't (yet) do (yet).