[Geoserver-devel] Versioning feedback - branches

Looking here: http://docs.codehaus.org/display/GEOS/Versioning+WFS+Requirements
I have a question about versioning and branches.
Mostly because I have never seen pure data being branched. Is branching something that will be needed or wanted with the versioning WFS?
From my experience if you want a copy of the data, you just copy it and it is just that, a 'copy'. Branch seems to imply that there will be a future merger, or a bit more of a relationship between copies of the dataset. Are we trying to inflict software development practices onto data users?
Yeh I see that ArcSDE supports the notion of branches. But I just don't get it. Is this all to do with holding areas for data validation? Someone enlighten me.

ps. Just throwing out some thoughts to spark discussion.

--
Brent Owens
(The Open Planning Project)

Brent Owens ha scritto:

Looking here: http://docs.codehaus.org/display/GEOS/Versioning+WFS+Requirements
I have a question about versioning and branches.
Mostly because I have never seen pure data being branched. Is branching something that will be needed or wanted with the versioning WFS?
From my experience if you want a copy of the data, you just copy it and it is just that, a 'copy'. Branch seems to imply that there will be a future merger, or a bit more of a relationship between copies of the dataset. Are we trying to inflict software development practices onto data users?

Yes, I'm evil and I want them to suffer just like us in our branching hell, he he he he :slight_smile:
No, seriously, branching pops up in all "enterprise" versioning system
for a variety of reasons:
* someone needs to work on a modification of the map, but keeping the
   main map untouched, eventually, but not necessarily, applying the
   changes to the main map;
* maybe there are multiple people doing so, exploring design
   alternatives or simply updating different areas, again with the need
   to make everything disappear if it turns out they did something wrong.

You say, a copy is a copy. Yeah, a 2GB worth of data copy, is another 2GB (no branching, no cheap copies). Plus, when you want to
merge work done by multiple people, you need to spot what changed and
eventually fire up a conflict: if you kept everything in the
version control you can, if you did not, you can again, but it's a lot
slower and more complex since you first have to find what changed (having the copy be versioned as well helps, but not as much as having
everything work against the same revision numbers).

Anyways, I think the data size argument is what makes a lot of the
difference.

Yeh I see that ArcSDE supports the notion of branches. But I just don't get it. Is this all to do with holding areas for data validation?

Yeah, it may have to do with human validation before merger.
But if you want to see a real world use case that not even ArcSde was
able to manage properly, see this:
http://gis.esri.com/library/userconf/proc01/professional/papers/pap232/p232.htm
(the only versioning system I know that can make modification on trunk flow automatically on open branches is Oracle Workspace Manager).

Cheers
Andrea