couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yuval Kogman <nothingm...@woobling.org>
Subject Re: reiterating transactions vs. replication
Date Mon, 25 May 2009 07:21:34 GMT
2009/5/25 Chris Anderson <jchris@apache.org>:

> Right, but at least in those cases I can diff the document to figure
> out what the conflict is. The bulk transactions you describe could
> become conflicting and before I could save the doc I'm working on, I'd
> have to figure which other doc was causing the conflict.

The idea is that you can ask for a conflict earlier if that's what's
going to help. If you need more context then arguably there should be
a way to ask for it to not be thrown away.

If you have a conflict in the aforementioned sorted collections and/or
graph nodes you need to know the state of other documents to merge
correctly.

I still think this is out of scope anyway. For a proper merge you'd
want the common ancestor documents as well, not just the symmetric
difference. This is best done by implementing a versioning model on
top of couchdb. But to implement this model consistently you arguably
still need atomic primitives.

> I'm not sure
> why not to just call the larger unit of data a single document, if
> that's how you want to use it.

So basically instead of using multiple documents all of my data would
go in one document? Why didn't I think of that ;-)

Mime
View raw message