Date: Thu, 25 Nov 93 12:09:37 +0100
From: Tim Berners-Lee <timbl@www3.cern.ch>
Message-Id: <9311251109.AA04241@www3.cern.ch>
To: mitra@path.net (Mitra)
Subject: news: and nntp: URLs
> Alan Emtage (bajan@bunyip.com) wrote:
> > : [Erik Ostrom writes:]
> : > Alan Emtage (bajan@bunyip.com) wrote:
> : > > - Group decided that the "news" URL should be removed
from the
> : > > document, and replaced with a valid NNTP: URL.
Installed base will
> : > > continue to use "news" URL until transition can be
made
> : > What will the nntp: URL contain? Will server names be
permitted? Allowed?
> : I think Mitra is the best person to handle this one....
>
> My understanding was that the URL would be sufficient to retrieve a
> message from a nntp server by LOCATION. Therefore that it would be
- in
> URL-bnf:
>
> nntpaddress n n t p : // host / group [ / digits ]
I FEEL THAT WE ARE LOSING SIGHT OF OUR OBJECTIVES AND DISAPPEARING
INTO A CLOUD OF INVENTION.
The URLs are NOT here as an invention -- apart form the universality
of the superset namespace there is nothing new. That is, the URL
is NOT here to change the way proctols are used.
The news: namespace was called news: because it is in fact one
of Karen's namespaces which do NOT imply an access protocol in fact.
In the Internet context, NNTP is implied. But the Usenet architecture
allows many protocols to be used -- and preseve the essentials
of message-id and newsgroup name. These are things which make
the news system hang together. NNTP with its article numbers
and its IP addresses is a newcomer and is NOT the basic namespace.
If you look at an NNTP: namespace using //host/group/number
you will find that it is very poor.
1. In REAL USE, nntp servers are NOT globally acceesible, and
to make them so would be an abuse of NNTP. For that you use
oetr things. So you can't use these URLs in any sort
of reference except between people using the same
new server.
2. An article may have many group/number paths as it is posted
to many groups
3. As we know, the numbers are different on different hosts
4. You can call it an access algorithm, but you have no way
of deriving this nntp: url from any other sort of name
such as the message-id esxcept by retrieving the message
[HEAD] using its message-id.
The advantage of this URL is that it will work with some common
broken NNTP implementations, which do not allow lookup of news:
URLs. I say that fixing the news servers is better than
damaging the whole NIR scene with rotten "nntp" URLs.
The argument -- at which I regret not having been present --
seems to have been along the lines:
1. "nntp" represents better what the "news" URL is.
2. As it's now called "nntp" it ought to be something different.
It is this fearsome urge to create taxonomies to constrain things
which are different into boxes of nominal similarity which is
preventing progress. Theer was a feeling I think (Mitra?) that
news: URIs were not locator-enough in feel, and ought to be
reduced in level. This stems from the URL/URN taxonomy and
is not useful. news: urls are what they are. They are things
which ACTUALLY APPEAR in the Newsgroup: and references: lines
of news messages. (I am not iunterested in URls for
.newsrc files which are the only place the group/num things
appear).
If you want to know the properties of a news:URL, look at the
usenet news specs _and usage_. Don't declare what they are and
are not -- it is better to spend one'd time declaring what angels
or unicorns are and are not.
We (the URI WG) are not inventing things with the URL stuff.
We are documenting existing practice. The inventions go into the new
schemes, not the old.
Tim Berners-Lee
CERN