[ckan-dev] Distributed changeset model for CKAN

Sun May 15 13:55:50 UTC 2011

Just to be absolutely clear: I said at the start of the mail that I
was specifically *not* suggesting we start intermingling distributed
stuff with the 'pending' discussion :-) but that people might not be
aware of the extensive existing work we did and some useful concepts
that came out of it :-)

On 14 May 2011 17:33, james at 3aims.com <james at 3aims.com> wrote:
> Good point, also worth noting that the already implemented CKAN to CKAN
> harvesting provides another approach for a distributed network of CKANs.

I don't know much about this so look forward to finding out more.

One question I'd if this is planned to used in full distributed model
(rather than a simple 'pull' model) is how you deal with edits to same
objects in different places (the merge problem). E.g.

Repo A (obj X):  revA1 -> revA2 -----------> revA3
                                         \                         \
Repo B (obj X):                   revB1 -> revB2 ---> revB3 (!conflict or merge)

> So as not to over engineer, or risk delaying revisioning implementation lets
> try to keep revisioning and distribution as separte discussions for now?

I thought that was exactly what I said in the original email :-) To
even clearer let's leave this thread here for the time being and
refocus all attention on the pending revisions' one :-)

Rufus