Allowing annotations at any place in the document is a level of complexity
higher than I had envisioned, and I'm not sure that it's completely
necessary for this. I definitely don't think the annotations should be
directly inserted into the documents. This is a question for the
user-interface folks out there - how could this best be implemented such
that the original document maintains its original state yet you can see
where people annotated? How then does this get implemented into HTTP?
I guess one answer is that when a document is retreived, an optional
comment-map is passed along, and the client users have the option of
seeing the document in its original state or marked up with hyperlinks
representing regions of the text. This comment-map would be dynamically
generated from the database of comment thread heads as known by the server.
I.e., the client gets
http://www.wired.com/path/file.html
and then when the author wants to see the comments
http://www.wired.com/cgi-bin/gencommentmap/path/file.html
One thing I don't like about this model is that it encourages replying to
only parts of the document rather than considering the document as a whole.
This is a problem on usenet - cascades and flame wars get vicious because
people liberally quote out of context and isolate certain phrases to
sound like the author's a Nazi sympathizer. :)
This also requires a higher level of complexity in the browsers.
Might I suggest we make this a later goal? That for now we allow as
part of the protocal the ability to map a comment to a location in
the file, but don't demand that from browsers and servers as yet?
> In his view, this whole thing is a store and forward system. It depends on a
> nano-tech storage system, so everyone would keep a complete copy of the Web
> with them at all times. Each Web module could plug into any other and upload
> and download new links to other module.
>
> Of course, this last idea is a bit fanciful, (if not impossible? What kind of
> bandwidth would be needed to download the entire Web in a few seconds?), but
> everything before that sounds like the Web we Know and Love today.
The above prophecy is, of course, USENET. Now, we strongly considered
simply opening up USENET newsgroups for what we wanted to do, but it
seemed impractical - i.e. wired.flux, wired.fetish, wired.net.surf, etc...
I wasn't convinced that all sites would be willing to carry us, nor that
we'd be able to control the quality and interface to the discussion as
much as we'd want. The WIT model (http://info.cern.ch/wit) is much more
like it - there is a LOT of value to a structured discussion, one that has
some sense of permanence, and a lot of value in "gardening" as we've seen
on the Well, and as such USENET doesn't (and by its nature probably never
will).
Brian