Re: Proposal: WIT over USENET or Mail

Tim Berners-Lee <timbl@www3.cern.ch>
Errors-To: listmaster@www0.cern.ch
Date: Wed, 15 Jun 1994 11:16:36 +0200
Errors-To: listmaster@www0.cern.ch
Message-id: <9406150912.AA02148@www3.cern.ch>
Errors-To: listmaster@www0.cern.ch
Reply-To: timbl@www3.cern.ch
Originator: www-talk@info.cern.ch
Sender: www-talk@www0.cern.ch
Precedence: bulk
From: Tim Berners-Lee <timbl@www3.cern.ch>
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Re: Proposal: WIT over USENET or Mail
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas


> > I feel that the key question is whether we want to offer a smooth
> migration path for newsgroups, which are very popular and widely used,
> to the current architecture and capabilities of the Internet (recall
> that NNTP was invented in the UUCP days), or whether we want to abandon
> them in favour of our "brave new world".  I feel that if a migration
> path can be offered, then it should be offered.

I would go further -- that we want the brave new world to have NNTP.
(If all the new news were acecssed by HTTP the net would melt just as
it would if all the HTTP documents were served using NNTP!).  There is a need
in fact for a cache-cache protocol which migrates seamlessly betewen
the two extremes of HTTP and NNTP.  I just don't think that using NNTP
 as HTTP helps in that regard. For users, HTTP/NNTP/mail gateways already  
give migration paths
and ways to allow users at all levels to hook up with appropriate
levels of functionality.

(If you want to migrate from driving on the left to driving on the right,
driving in the middle is not a good strategy!)

> I must admit I wasn't aware that URLs of the form "news://host/article"
> were "forbidden";

They aren't in the spec.  They were proposed on the URI group and
were called "nntp:" not "news".

> to me they seem natural, self explanatory,
		  yes	   yes
> architecturally sound,
		  no -- as I tried to explain, you get messed up if you mix
		  these URIs with ordinary news as when you have a reference
		  in NNTP you don't know which sort it is.  This breaks the
		  system.  If I get a cross-reference in a local news article
		  I can't tell whether it is to a local or global article.
		  There may be other problems but that is one.
		  

> and useful for newsgroups whose ownsers don't
> wish to flood the entire Internet, yet wish to allow access to anyone
> interested in participating.  So "forbidding" them in the name of some
> higher principles ("though shall abandon NNTP in favour of HTTP?" :-)
> seems to me to be a mistake.

The higher principles are *not* that that one should migrate, but that
one should use NNTP for broadcast and HTTP for individual retrieval,
and that one should have a consistent model of reference before
defining new URIs.

> -- Udi

Tim