Re: [Storagesync] Storagesync Digest, Vol 5, Issue 1

[Ted Lemon <[e-mail]>]

Wednesday, Dec 9, 2015 4:13 PM Klaas Freitag wrote:
> Automatically mergeable or not, I don't think resolving conflicts shoudl
> be part of the sync protocol. Of course conflicts must be detected and
> hooks should exist to plug in code to work with it, but I doubt it
> should be part of the sync layer to interpret and resolve them.

I agree, as I said to Marc.   But the sync protocol has to _support_ the accurate detection of conflicts, or there's no way to resolve them, because you don't know about them (until your change has been overwritten, at which point it's too late).

> And even if there is an automatic resolving of conflicts for some file
> types (that is obviously possible if that is wanted from a higher level
> point of view) there always remains the possibility of not at all
> automatically mergeable conflicts.

Yes, I think we all agree that this is a problem.   Marc mentioned that he thinks it's a research problem.   I think a general solution is a research problem, and not a very interesting one: all I really want is a way to know that the conflict exists and what the sequence of events was; if an automatic merge is possible, that's great, but not required.

> I am also not sure about the importance of the sequence of changes, as
> there are situations caused for example by offline work, where changes
> happen in parallel, these cause a conflict, and it's not really
> important which change was before or after. Important is that both
> versions can be found again, but not necessarily in timely order.

Not remembering the order in which the changes occurred is an optimization which gains you almost no efficiency and costs you greatly in terms of functionality.

> I do not understand why Kuba said "Etags are sufficient" in this
> context, but maybe I lack context because I am late to the party...

Etags are sufficient if you just want to see if what you have is more recent in time than what's on the server.   They don't allow you to detect interleaved updates.


--
Sent from Whiteout Mail - https://whiteout.io

My PGP key: https://keys.whiteout.io/[e-mail]