I think in terms of working out the architecture for YNMP to handle
future elections, things should probably just be factored as we go,
though what you suggest is reasonable - though I think any user
interface is going to have to be heavily denormalised for the user's sake :)
Another one to add to that list is the proposed PR for the House of
Lords. Actually I think it'd be really interesting to think how we'd
handle an elected house of lords - the candidate centric approach we've
done this election would be quite applicable (praying that it won't be
party list, one can hope, right?).
As for STV I've got no idea how other organisations tend to release STV
votes, but in our student union elections they release the
round-by-round vote assignments (and also actually release all the
ballots, anonymized). Visualising this is pretty tricky! I could imagine
how you could show the losing candidates's votes splitting into the
other candidates.
-t
On 12/05/10 12:45, Edmund von der Burg wrote:
> Exactly - it is a nasty rabbit hole that I'm wary of going down.
> Another example: The general election was on May 6th but not all
> results are in because one of the candidates died so that poll has
> been delayed until May 27th.
>
>
http://www.yorkpress.co.uk/news/8115410.Thirsk_and_Malton_election_postponed_after_candidate_dies/
>
> So this means that there needs to be a table for elections (eg
> 'General Election 2010' ) which links out to a table for contests (
> election: 'GE20101', date: '27 May 2010', place: 'Thirsk and Maldon
> Westminter Constituency (2010)' ) which in turn links to the
> candidates contesting it (candidate_id, party_id, votes) which finally
> gets us to the candidate.
>
> Not forgetting that candidates seem terribly keen on standing under
> two parties ( eq lab& co-op ) or better yet making shit up (one