Annotation support

Marc Andreessen (
Sat, 10 Jul 93 03:53:22 -0500

David Martland writes:
> Support for Annotations is not (as far as I know) part of the
> current HTML standard, since it is not really a property of the
> source text, but rather of the Browser.

Yup. The URL (and, in the future, URN) mechanism provides a
convenient infrastructure for cleanly layering annotations on top of
documents at any of several virtual levels, and we're exploiting it.

> I notice that NCSA's Mosaic is currently providing private
> annotations, and is about to provide group and public annotations.
> This is a good idea, although the annotations are on a "per Window"
> basis, which may not be ideal - it could be better if the
> annotations are linked more closely to sections of text - perhaps by
> having "annotation anchors" in the margin, which link to the
> annotation text

Absolutely. The reason we don't do this yet is this -- how do we
track position in a text document that (a) we have no control over and
(b) could change in any way at any time? I don't have any answers
yet, so I'm not sure how to work it...

(If we don't come up with anything better, ultimately Mosaic will use
a combination of character position plus some amount of textual
context, with a fallback mechanism to drop "detached" annotations to
the bottom of the document with clear notification that this has

The same mechanism, whatever it is, will also be used to support
inlined hyperlinks in the future, btw.

> There is an additional problem with annotations, not addressed in
> Mosaic's implementation, since it may be desirable to allow
> annotations on annotations! I do not believe that Mosaic does
> this.

As you noted later, Mosaic sure does...

> I am aware of some systems which provide support for annotations,
> though not for HTML documents - for example the MUCH system
> developed by Roy Rada at Liverpool, and designed for asynchronous
> local development of text, does have annotation support.
> However, although it might appear that annotations are "simply" a
> property of the Browser - the problem does not end there. The way
> NCSA have done it is to provide annotation servers - this then
> raises the possibility of different Browsers using the same
> annotation server. This would be beneficial to many users, since
> there would then be no requirement to adopt a particular browser
> package - for example I could communicate using Mosaic on my Unix
> system, and someone else could annotate my text using Cello on an
> IBM PC etc.

Yup, and in fact Bill Perry is currently adding annotation support to
his w3.el Emacs client, and Mac and Windows Mosaic will have
annotation server support (it's already working under Windows, as a
matter of fact, so I can *guarantee* that :-).

> It seems to me that there is a need for consideration of annotation
> services, and how to provide them.
> I think this is an issue which needs wider discussion - if it has
> not had it already.

Absolutely! I'm looking forward to hearing what you and others have
to say. The current system is only a prototype, and we are absolutely
open to suggestions (or completely redesigns). See
for a few notes I've typed up...


Marc Andreessen
Software Development Group
National Center for Supercomputing Applications