On Sat, Jun 06, 2020 at 11:59:47AM +0200, Thibaut Verron wrote:
> >
> > What I think is missing is not the description of a specific protocol,
> > but a higher-level spec of basic capabilities needed for the
> > collaborative editing support in Emacs.  Is this available anywhere?
> > If not, could someone please write it up?
> >
> > For example, one thing that strikes me is why "collaboration" via a
> > dVCS is not a good solution, or at least the basis of a solution?  Am
> > I missing something?
> >
> 
> Collaborative editors usually show modifications done by other users in
> real time. I don't know how major conflicts are resolved.

My hunch is that conflicts aren't as heavy "in real time", because users
can react to them in a more fine-grained fashion. But there's a whole
body of theory dedicated to that. I'd start here [1].

Actually the problems are akin to (but possibly not as hard as)
networked gaming, where you have several clients sharing a common
model: you'll have to cheat a bit if you don't want to wait until
you know the exact model's state, because that'd mean a full network
round trip. Sometimes you can afford that, sometimes not. Balancing
out that and fixing things to hide your cheating after the fact
in a way that the user can cope with the fallout is, I think, the
"interesting" part.

> How would you emulate this with a VCS? Commit-push-pull with various
> --force flags, on a timer run every second?

I think network latency would kill you (or rather, your users might ;-)

Cheers

[1] https://en.wikipedia.org/wiki/Collaborative_real-time_editor#Technical_challenges
-- tomás