Re Reading news via WWW browsers

Dave_Raggett <dsr@hplb.hpl.hp.com>
From: Dave_Raggett <dsr@hplb.hpl.hp.com>
Message-id: <9303011813.AA12988@manuel.hpl.hp.com>
Subject: Re Reading news via WWW browsers
To: www-talk@nxoc01.cern.ch
Date: Mon, 1 Mar 93 18:13:11 GMT
Mailer: Elm [revision: 66.36.1.1]
> Most discussions in most newsgroups consist of threads. Someone starts a
> thread, others respond, and so forth. We are using here a newsreader called
> TIN (written by Iain.Lea%anl433.uucp@Germany.EU.net) that whenever a news
> group in entered, it first indexes all the postings into threads, and then
> presents a view of the group traffic as a collection of threads.
> You pick a thread, and start from the first posting. The next posting you
> will get will be the next one along the thread.

Yes, most of us here use the Notes interface, which maintains such threads
or "notes strings" as part of its database. The current NNTP news servers are
much too dumb for this - it takes far too many messages for the client to
work out the threads. You can do it, but it is painfully slow if your
NNTP server is on a remote machine.

> This approach lends itself very well to a hypertext browser. Using the 
> same indexing method as TIN, it will embed a link into the end of  each 
> article that will point to the next one in the thread, in addition to
> links in the beginning of the article that point to the previous article
> in the thread and to the start of the thread.

Yes, I too have been thinking along these lines. The first step is an "agent"
program which runs daily to get the complete list of news groups, sorts them
alphabetically then builds them into a small group of html docs that allow
users to browse the newsgroup names hierarchically. The next step is to
extend the agent to build an html doc for each newsgroup's index. This
involves getting the headers for any new articles in that newsgroup and
can be linked to the job of tracking backward references, and updating a
database of threads/notes strings.

In the longer term, we can lobby for such support to be added into future
releases of the NNTP daemon. Our only hope in the short term is to build
agent processes that do the additional processing as a regular background
activity. You can think of these as a variety of "knowbots" which carry out
tasks that are impractical to perform on an interactive basis.

I am sure NEWS is due for a shakeup what with the arrival of MIME and html as
a registered format. The current restrictions on character sets and plain
text is rather cramping these days. Furthermore the number of newsgroups has
grown enormously over the last few years - the ability to carry popular
groups locally and to get the others on request from remote servers seems
long overdue.

> Is there a way to read news groups not carried locally via the WWW ?
> We have here a rather limited selection on newsgroups (only 1500 or so :-)
> Some people are not that fortunate, and I know of many groups we do not
> carry either. Additionally, postings are kept here only for about 3 days.

I have been toying with an extension to the URL syntax for just this purpose:

        news://news.ysu.edu/alt.culture.*

This implies the list of newsgroups under alt.culture on the NNTP server
"news.ysu.edu".

        news://news.ysu.edu/alt.culture.usenet

This example specifies the newsgroup "alt.culture.usenet" as held on the
nntp server  "news.ysu.edu"

        news://news.ysu.edu/<C34u1w.7Ls@news.ysu.edu>

This specifies an actual news article from that newsgroup and server.

One outstanding problem is which machines archive what newsgroups? It is
often the case that the original article has been discarded from the local
server, and it would be really great if the browser could then try to get the
missing article from a remote archive system. The establishment of well known
servers for given groups would be a good start.

Best wishes

Dave Raggett, HPLabs, Bristol, England