[kforge-dev] on versioning

Mon Dec 4 19:41:41 UTC 2006

dear all,

Not sure if this is the right list to linkdump the mentions from 
earlier versioning-related conversation with Rufus, as it's not really
kforge-specific but is too specific for okfn-discuss; but here goes.

Andrea Aime of GeoServer has been working on adding versions to the
transaction mode of the Web Feature Service protocol (XML exchange of
vector data). This may sound totally irrelevant but as a byproduct he
did a lot of surveying and stresstesting of versioning techniques.

http://docs.codehaus.org/display/GEOS/Versioning+WFS
particularly of interest probably are:
http://docs.codehaus.org/display/GEOS/Versioning+WFS+-+Versioning+approaches+classification
http://docs.codehaus.org/display/GEOS/Versioning+WFS+-+Database+design
explanation of how OSM handles things now (oddly)
http://docs.codehaus.org/display/GEOS/Versioning+WFS+-+OpenStreetMap

I remembered this very rambling email i wrote about a 'reflecting'
store for keeping versions of RDF statements that handles the "if
object X merges with object Y the history of whichever one gets
deleted points to an object that no longer exists" problem. (which is
a really common case in an aggregating RDF stores)
http://lists.usefulinc.com/pipermail/redland-dev/2005-December/001222.html
Of course i never finished 'reflect', it was a semi-silly idea and i
never had a realworld use case for it.

Hope any of this helps; i am on the lookout for a sensible way to
replace ZODB's 'long transactions' versioning approach which is
currently in-flux broken, and would benefit from thinking harder
about all this; i seem to be having a versioning conversation every
second day, recently. 

cheers,

jo 