W3C home > Mailing lists > Public > www-annotation@w3.org > July to December 1996

Re: Does annotation need a huge central data base?

From: Misha Glouberman <misha@the-wire.com>
Date: Mon, 12 Aug 1996 23:45:33 -0400
Message-Id: <>
To: www-annotation@w3.org
Jacob Palme writes:

>An annotation system does not need a large central data base.
>If you want to annotate a document stored at www.foo.bar.net,
>then first check if www.foo.bar.net has an annotation server. If so,
>store it there. If www.foo.bar.net does not have any annotation
>server, check if foo.bar.net has one, and store it there. If
>foo.bar.net has no annotation server, check for a server at bar.net.

I like the idea of splitting reponsibility by url.

I'm curious about a couple of things: 
- What would motivate sites to allocate resources for hosting their own
comments, given that they'd know that if they didn't, someone else would
pick up the slack for them? 
- If lots of sites (even a sizable minority) chose not to host their own
comments, how would the central servers (.net, .com, etc..) get funded?

And Daniel LaLiberte writes:

>I believe the issue Misha is addressing by suggesting a central data base
>is one of control.  The publically available annotations of a document
>might not be compatible with the policies of the server of the
>document; the administrators might want to suppress the annotations in
>fact.  So these annotations would either not be available or must be
>available through some administratively separate server.

Thanks. Exactly my point.

>Now the problem for the user who wishes to see the annotations is how
>to find this separate server.  So there are problems either way: a
>problem of control by the servers of documents being annotated and, if
>not that, a problem of finding the annotations.

Hmm. I think Jacob's got a pretty good answer on how to find the separate
server. In principle, the system could easily be refined further, with a
single site taking care of a*.com, another for b*.com, and so on. 

Seems to me that the real problem in the kind of system that both Jacob and
I are describing (ie- one where you can see _all_ the comments related to
_any_ page on the web, versus one where you only see comments in the sets
you're subscribed to) is that you're talking about a lot of storage, and a
lot of requests. You can propose a system with multiple servers (as Jacob
does), but that might just distribute the problem more broadly rather than
solving it.

>A few centralized databases for all annotations of all documents would
>solve these problems, but it would create new problems.  Control by
>the centralized databases is obviously a problem.  

Say the databases were centralized but independent of the servers being
commented upon. Would it be possible for the comments to be encrypted so
that the system administrators couldn't alter or remove individual posts?
(This isn't a rhetorical question- I'm not as familiar with security as I'd
like, and would be curious as to what headaches this would create).

>The scaling problem
>is also very severe.


That's why I think the really critical and interesting part of designing a
universal-comments system is figuring out how to create motivations
(financial or otherwise) for comment-hosts. My paper touches that issue- I
plan to develop it  more.

>I think the solution has to be many more annotation servers focused on
>much smaller user groups and smaller sets of documents. 

has to be?

        - Misha
Received on Monday, 12 August 1996 23:47:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:16:53 UTC