Re: WIT

hallam@dxal18.cern.ch (HALLAM-BAKER Phillip)
Errors-To: listmaster@www0.cern.ch
Date: Sat, 11 Jun 1994 18:44:41 +0200
Errors-To: listmaster@www0.cern.ch
Message-id: <9406111643.AA26854@dxal18.cern.ch>
Errors-To: listmaster@www0.cern.ch
Reply-To: hallam@dxal18.cern.ch
Originator: www-talk@info.cern.ch
Sender: www-talk@www0.cern.ch
Precedence: bulk
From: hallam@dxal18.cern.ch (HALLAM-BAKER Phillip)
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Re: WIT
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas

In article <73EA@cernvm.cern.ch> you write:

|>    Date: Fri, 10 Jun 1994 12:37 EDT
|>    From: hallam@dxal18.cern.ch (HALLAM-BAKER Phillip)
|>
|>    I think we need to try to focus on what the exact nature of the
|>    problems of WIT are and generaly try to get a bit more constructive.
|>    I don't think that anyone has suggested that it is perfect. Until
|>    you have something to play with it is difficult to know what you want.
|>
|>Right on the money.  Rome wasn't built in day.

<DISAGRE>
Rome was built in a day by sewing the teeth of the Hydra from which sprang
a army who built Rome. For some reason the Hydras teeth only grew men so
the Romans then went off to be very politicaly incorrect to the Sabines.


|>    I had a go at an email hypertexter and gateway. Also properly hypertexting
|>    news. Both are a big win in themselves but what is still missing is a
|>    framework for discussion.
|>
|>Can these be interfaced to WIT so that people can comment on WWW-talk and
|>friends?

<RESPONSE>

We are working on interfacing them to libwww. But these things take time.
Rome Wasn't built in a day you know.

|>
|>    It is quite easy to imagine a set of conventions that could interface
|>    mail and news to WIT. In fact they are the same thing if you think about
|>    it. News is simply an email with 20 million recipients, most of whome will
|>    file it as junk and delete unread.
|>
|>The main difference is the propagation pattern.
|>
|>There remains the issue of maintaining copies of messages.  With WWW, one
|>could progate URLs (URNs) and headers via NNTP, and users/sites could resolve
|>and cache them, according to the relevance for their site.

<AGREE>

I like the idea of taking the NNTP model and munging it. Perhaps what we
should try to do is to layer the system. We have two very separate problems:-

<PROPOSE>
1) The structure of News is unwieldy and will become more so with more	
	users. Currently we are about the size of Luxembourg. Within five 
	years we should be the size of Great Britain.

	The bigger the city, the more drunks you find in the pubs.

2) The distribution of news is wastefull. Every site gets everything. 
	What I would like to do is to prune down the NNTP system to a
	two stage system:-

	1) An NNTP backbone 
	2) Cache Relays hanging off the backbone.


	Starting from the network topology we cut down the network to
	a logical system of sites, minimizing the duplicate paths etc.
	Main hub sites of the internet such as CERN, MIT etc would 
	carry the full service.

	The sub-sites would run stripped down service. They would
	typicaly subscribe to a number of sites and for the others
	take "headers only" service. Instead of getting the bodies
	they would get only heads. These could be compressed to save
	even more space.

	When a sub site gets a request for a message not in the cache
	it requests and caches it.

<MY-RIGHT-ANGLE-BRACKET-KEY-HAS-MELTED-]

This system would be IMHO better for almost everyone. Users at leaf sites
would see a slight degredation in response. But there would be more leaf
sites. There would no longer be a need to run expiry in the braindamage
manner of USEnet. If an article has been flushed from the cache simply
ask for a re-send from the hub node. Hub nodes would keep archive systems
which would allow reference to *any* Usenet post *ever*. This would not be
as expensive as it may sound, USEnet on CD is avaliable. A jukebox is not
a prohibitively expensive item. Needless to say cataloging 
alt.binaries.dirty.gifs would not have to be a priority.

We only really need one long term store...


|>Can your interface handle MIME multipart messages?

Rome wasn't built in a day (RWBIAD)... but my MIME parser probably will be.

|>
|>WRT to NNTP as a transport, would it make sense to have an HTML/WWW hierarchy
|>where all the postings are marked up in html?

YES YES YES!!!!!

www. hierarchy here we come !!!!

Actualy I think we should create it anyway and gate it to the CERN/MIT link in
the same manner as the cern. hiearchy. The news software could be set to
allow world access to the www hierarchy only. Then we could get rid of these
!@#%&!@#^%#*$ mailing lists.



Before we get carried away though, I hope people realise that whatever our
intentions to create something of use to mankind it will only gain acceptance
when someone discovers how to use it to transport pornography. Perhaps 
<AGREE> and <DISAGREE> is the wrong model, perhaps we should have marks out 
of ten?

--
Phillip M. Hallam-Baker

Not Speaking for anyone else.