local editing: identifiers?

[Erik Ostrom <eostrom@research.att.com>]

dave and zephaniah wrote:
>> [ also, maybe we add a third message, s->c, mcp-edit-update, which
>> requests some kind of update on the text stored at the client? ]
>
>That gets nasty, because you really don't want to undo what the client has
>been working on for 3 hours because someone added one line of code..
>
>Perhaps a alert that it has changed and a way to request the changed text?

i think we should go with a two-message spec for now.  you're already
getting into tricky issues concerning persistence of identifiers,
whereas in the spec as it stands, it looks like an ident is in fact
not so much an identifier as something that the server can parse into
some useful notion of "location".

i said last time that i think the right solution to persistent
identifiers with updates etc is cords, and i still think that.  i
think the identifier you want pertains not to a location but to a
session or channel (where the session knows something about its
corresponding location on the server side), and i see no sense in
reinventing session IDs when we already have something that works fine
for that.

so anyway, i'd like to see a simple two-message system that we know is
stupid sometimes but that we can get multiple implementations of
without much effort.  and then we can start experimenting with
refinements on the scheme.

--erik