Concurrent Access

Ok,
 this is a fine question. What about when multiple simultaneous instances of
a document are operating? One crude but useful way to deal with this is for
every operator to have a clone of some baseline encoding of a document.
(the registered edition).

Then, when the personal extention to the clone is added it is saved with the
clone in the personal workspace. Only when the author requests replication
and exchange
with the shared document collection does the system worry about the
concurrent usage.

There is a class of problem in the real world where a human being can be
called upon to rectify the true essence of the active document.The
'coordinator' (cootor) of the document can request and formalize a
clarification extention to the shared document pool.
This is probably a common feature of many seemingly complex data
collections.

The rare collection will require fancy semaphores and 'extents' and
'boundaries' and we should bear these in mind, but we should not loose sight
of the principle that we are trying to create an idea which models a
requirement of real problems.

I know we are trying to be as implementation non-specific (ie. general) as
possible, but can we agree on a user model? Do users request an explicit
release and exchange? or are all edits supposed to shared instantaneously
(how useful is the latter in practice?)

I and my partners have had considerable experience with how users perceive
and use a web based resource, and we have found the following: (as predicted
(by others) at w3collab Boston) that effective groups of collaborators are
limited to 12 members. If a complex multi team problem is encountered, it
will be solved by 12 groups of 12, not by a team of 144.

If only we could come up with the 12 essential roles and responsibilities
for people.

We have a simplified implementation of an assignable role hypertext by
mail/FTP/composition system... and the lessons seem to be: base it on what
people actually do with a document, not what we'd do in a perfect world.
(see http://orca.fastwfd.com/samples for some samples)

If we are abstracting and encapsulating properly with our implementation,
then the users will be forced to do the right thing without even knowing it.
Short of getting their usernames and passwords right it should not be the
responsibility of the users to enforce all the constraints.

Remember the plan of Bush's Memex: let the users build, extend, and
replicate portions of the 'collected wisdom' so they can trade with each
other. Let's not obsess on what we might be able to accomplish to the
detriment of what the users really want.

In my experience, and contrary to popular thought, users do not want to be
publishers; users want to be authors. Publishing is an ill-appreciated art
which cannot die. Spelling will always be important for
professional products. Nobody cares about spelling; people care about
connotation.

Les Cuff
Freelance Consulant
Gods Country: Paradise, Newfoundland
lez@cuff.nf.ca.

Received on Monday, 27 July 1998 22:52:54 UTC