Date: Wed, 15 Jun 1994 19:46:44 +0200
Message-Id: <9406151746.AA11241@dxmint.cern.ch>
From: hallam@alws.cern.ch
Subject: Re: Proposal: WIT over USENET or Mail
>Why can't we resolve news URLs where the host name is omitted to mean
>"global" news articles, and look them up in the default NNTP server,
>and resolve full news URLs to mean the article at the specified host?
>This may require certain discipline in creating cross-references and
>other URLs based on the type of newsgroup.
I disagree with both sides:
Side A: News articles are avaliable by a flood fill system. A news server
should be able to access any article. Therefore there is no need
for a URL for news, a URN is sufficient.
Side B: Some servers have a closed set of newsgroups. Often only avaliable
at one site, sometimes at a highly restricted number of sites.
Thus there should be some method of specifying the host.
First off I would propose that we state our concerns:-
1) That when a user follows a link they get the data they want
2) That the optimum network transport is used.
3) That we do not constrain unnecessarily the setup of news systems.
4) That the URL specification has the highest possible degree of
symetry and consistency.
news://ptsun00.cern.ch:119/article@id
Fails the criteria if you
1) Have no permission to use ptsun
2) Are in the USA and there is a nearer site
Is based on the assumption that there is one news system. I don't think
that this is the case. There are serveral news hierarchies that are non
standard such as desy, cern, slac, vmsnet etc. Just because we can access
them all at CERN does not mean that everyone can. We have a URI but it is
not a URL. It fails to locate when it could easily do so.
This format provides the newsgroup as well as the article id. It is a suitable
format to POST to. It is also usefull for a GET. If the browser does not need
the newsgroup info to find the article then it can be ignored. The original
format can also be used in this case. But the inclusion of the newsgroup does
allow access to more than one NNTP server.
NNTP is a flood fill system. It works as such. If I want to transport HTML
documents using flood fill I should be able to chose to use it *WITHOUT*
having to involve USEnet. For example I would like to be able to keep
three or four mirror sites for documentation going using a flood fill system
Why not use NNTP?
Another point to consider that might make all this moot. If you used a proxy
server to read and post news that server could perform intelligent
hypetexting of messages. Why not permit the setup of that *server* to be
capable of accessing the newsgroups of more than one NNTP system. The
server would be doing the caching etc, etc. It would be sensible for
such a server to be able to reference articles by host within the cache
and for all sorts of internal reasons. This is where I think that there is
a case for news://ptsun000.cern.ch:119/.
Such an architecture (NNTP backbone with intelligent cacheing HTTP modules
hanging off it) could serve as a more extensible model for the USEnet
system. As Usenet grows the proportion of data we are sending and storing in
flood fill is getting silly.
Phill H-B