[Rebuild] Tests of rebuild process - we need your input and help

[Dermot McNally]

Hi Simon,

It's a useful question on a fairly late-breaking part of the rebuild
process. So whereas most of what's planned has been understood for a
while now (even the exceptional changesets), we do need to work out
where in the process we ought to handle exceptions that apply at the
level of a single object.

Comments inline...

On 23 March 2012 09:50, Simon Poole <simon at poole.ch> wrote:

> Not so much a technical question, but one of policy: are we  going to
> throw out individual CT acceptance then for all of PL?

I don't think we can, mostly because it's hard to define what that
means. Would we negate CT for anybody ever to have mapped anything in
Poland? Would we expect a majority of edits there? An "overwhelming
majority", perhaps?

In any case, the only problematic case is a mapper who has performed a
UMP import. Using cheerfully invented statistics I will boldly claim
that only a subset of mappers ever does imports. Any action we might
take to reject data from a mapper having agreed to ODbL in good faith
would IMHO have to be at the level of identified problem data, at
least at first. If we identified a systematic copyright infringer, for
instance, there's a good chance we would wish to roll back all edits.
This has of course happened in the past and can happen again
independently of the licence change.

> I would have
> assumed that we would have only considered an object clean/dirty list
> for such accounts that had accepted the CTs. The whole point of the
> effort being to allow contributors to agree even though they have
> tainted data in their contributions and not to maximize the amount of
> data we can import from UMP.

Ah, are you considering the case of a mapper not agreeing to CT but
where (s)he contributed objects that UMP will declare clean? This is
an interesting case, not least because we have had issues before where
otherwise non-agreeing mappers will have data retained that were
mechanically imported or manipulated (often with the direct
involvement of the mapper).

You are right that this is a matter of policy, but it's also one of
co-ordination. For instance, we are told that the Polish community is
collaborating with UMP to produce a list of objects that can be
declared clean for ODbL. If our rebuild process will recognise somehow
(and this is why I tend to favour a dirty object list rather than a
clean object list) a UMP object that cannot be retained, this would
legitimise the agreement of those Polish mappers who have agreed
despite having imported UMP data (assuming this has happened).
Additionally, it would allow for Polish mappers who have so far not
agreed (because of their incompatible imports) to agree now. This is
one possible policy that could work.

Frederik, have you been close enough to the Polish efforts to
understand more about this?

Dermot