Differential Synchronization vs OT

[David Hjelle]

I've been reading up a lot on various ways to create a realtime, collaborative, web-based environment, and really appreciated the earlier post on CRDTs vs OT. I was wondering if anyone had any experience with Neil Fraser's "Differential Synchronization" approach as compared to either OT or CRDT?

Thanks for any insights anyone may have!

[Joseph Gentle]

No, but I'm always very skeptical of diff-match-patch algorithms.
They're usually not stable / reliable in the face of conflicts.
Diffing is also expensive and unnecessary if you know what changes
you're making to the data structure anyway.

Reading between the lines in the intro of that paper, its saying that
diff-match-patch is useful just so you don't have to make many changes
to existing systems, and you can hack on diffs. But ShareJS isn't
designed with that kind of legacy in mind.

-J

> --
> You received this message because you are subscribed to the Google Groups
> "ShareJS" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sharejs+u...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[David Hjelle]

Thanks for your input! Your skepticism matches what was rather my final (admittedly subjective) conclusion: you can demonstrate nice properties about operational transformation and about CRDTs, but it gets a lot harder to prove anything about diff-match-patch algorithms.

Thanks again!