Re: support for setext, the structure-enhanced text format, in WWW

marca@ncsa.uiuc.edu (Marc Andreessen)
Date: Tue, 4 May 93 02:33:02 -0500
From: marca@ncsa.uiuc.edu (Marc Andreessen)
Message-id: <9305040733.AA03541@wintermute.ncsa.uiuc.edu>
To: pflynn@curia.ucc.ie (Peter Flynn)
Cc: www-talk@nxoc01.cern.ch
Subject: Re: support for setext, the structure-enhanced text format, in WWW
In-reply-to: <9305031003.AA05900@curia.ucc.ie>
References: <9305031003.AA05900@curia.ucc.ie>
X-Md4-Signature: b1782bfca1d5e54f8cc8633c8b8b16df

The upshot of this is that this is the same problem I sent mail to
www-talk about a few days ago.  Many news servers do not honor the RFC
977 (NNTP protocol spec) requirement to support fetching articles by
article ID -- they only allow fetching by group and message number.

This is a big, big problem.  There really aren't any answers, except
to have one's news admin switch to a working NNTP server (I'm not sure
which versions of what are working and broken, only that some servers
work and some don't).  URL's aren't currently designed to reference
articles by group and number, and as Tim pointed out this isn't really
feasible, since message number is a very transient property.

If anybody has any other ideas, please post them to the list....

Cheers,
Marc


Peter Flynn writes:
> I have found a problem in retrieving news thru both www (linemode)
> and xmosaic. Both correctly pick up my news server name, and if I
> follow the links from `By Type' thru `Network News' to a specific
> group, both programs correctly list the articles, but cannot access
> the article texts, giving me instead the message `Connecting to
> NewsHost ...' and then coming back with `Back, Quit or Help:'
> 
> It is clear that both programs are expecting articles to be stored
> in files with names set to the message-id. I have never encountered
> a system which does this, and my own server runs plain vanilla nntp
> and Cnews, which puts articles in files whose names are the article
> numbers (sequential, 1+).
> 
> As the www server is capable of finding the directory (and is
> presumably grepping for 'Subject:' and piping the results back into
> HTML---or is it actually opening each file in turn?) could not a
> simple addition to the code see if the files are named (=message-ID)
> or numbered (=message-number) and act accordingly?
> 
> If this is not possible, where do I find a news system which stores
> articles under message-ID filenames instead?