Re: Back-link annotation set scenerio


> My first thought on the back-link idea was that this should be handled
> with a different protocol that specializes in links.  There is an Atlas
> protocol developed at GaTech that does this (authors on the cc list). 
> This would handle forward links as well as back-links, and it would
> probably be managed in cooperation with the server of the documents
> that it manages links for.

True, Atlas is your basic hyperlink database, but with HTTP-ish commands
to query the server for topology.

> But then my second thought was why does it make sense to do it with
> annotations? Certainly annotations can handle links too, but another
> way to look at the problem is that links can handle annotations too. 
> In fact, the link people think that way.  Who wins?  I don't know yet.
> Maybe they will turn out to be the same thing.

In our opinion, link databases can indeed handle annotations.  It'd
be nice to have better include statements in HTML for this though.

> 	But let's make truth-checking:  It is not gigantic;
> 	indeed we know these numbers (cf. http://www.webcrawler.com/). The
> 	top five are ...
> I think Wayne was thinking of a single back-link server to handle
> all back-links in the whole world.  But I don't think there would
> be a need to concentrate all of those back-links in the same server,
> and if you tried, you'd have to somehow distribute it almost as soon,
> as we discussed.  Instead, each server, or each URL, could have its own
> back-link server associated with it.  These would be endorsed
> annotation servers, if it were done that way.  

A single back-link server is a really bad design decision, given the
previous research into centralized databases.  Essentailly, each WWW
server/repository needs a link database anyway for maintainence and 
visual management.  Indeed, the new products by Adobe's SiteMill and
Netscape's Liveworks do this.  Atlas provides a protocol for the 
rest of the world to utilize this information to faciliate visualization
and navigation in an effieient manner.  Annotation support could be added
to the protocol with no fundamental change in Atlas.

Jim Pitkow.

See the below URL for more on Atlas.