Annotation & asynch collaboration (Marc Andreessen)
Date: Fri, 20 Nov 92 21:00:54 -0800
From: (Marc Andreessen)
Message-id: <>
Subject: Annotation & asynch collaboration
I've patched httpd to allow remote annotations (an enabling mechanism
for asynchronous collaboration).  My httpd accepts the following

ANNOTATE /file/to/annotate?/file/of/annotation?Text describing annotation.

The result of this command is that /file/to/annotate gets a single line
appended to it; this line is exactly:

<A HREF=/file/of/annotation>Text describing annotation.</A> <P>\n


(1) '?' is used as the divider between ``keywords'', as it is with GET
    (for searches).  I'm content with overloading '?' (since it's
    already known to be safe, since it's not supposed to be part of a
    document ID), but other people might not be.  (If not '?', what?)
(2) /file/of/annotation can be any file with any access type, but is
    normally assumed to be a file residing on the annotator's local
    system and under her control.
(3) The new link can be arbitrarily named; it may be a good idea to
    have clients that implement annotation automatically prepend the
    annotator's full name to this field before sending it to httpd,
    but httpd shouldn't have to worry about this.

The security mechanism for this is as follows:

(1) A file to be annotated must ``pass'' the rules normally, and also:
(2) A new rule ``open'' is defined; it serves the same role (and has
    the same syntax) as ``pass'' except it determines whether a file
    can be annotated; and also:
(3) Since httpd is (or should be) normally run as 'nobody', the file
    must be writeable by 'nobody'.

This implementation does not cover informing the client whether or not
a file is ``open'' until an annotation is attempted.  That capability
should probably be addressed separately (??).

I'm sending the httpd patches to Tim tonight; unless there are
problems[1] with this approach, I'd love to see it implemented, and
I'll even set up the first annotatable web site...


[1] Dan will immediately notice that quotes aren't put around the HREF
field above.  At the moment, I'm going to run without them (since, for
one thing, I need to demo this with Midaswww on Monday), but the
official implementation should do the quotes.