[wdmmg-dev] Proposed model overhaul

Wed Oct 5 10:59:57 UTC 2011

On Wed, Oct 5, 2011 at 12:33 PM, Martin Keegan <martin.keegan at okfn.org> wrote:
>>> So, what will happen if I have a dataset which has columns which are
>>> obviously classifiers, but we don't know what taxonomy they're from,
>>> because it's a taxonomy only used in the City of Incognito, Alabama?
>>> They surely don't go in incognito_entity?
>>
>> Doesn't that question lose importance with the dataset split anyway?
>> This week it may be incognito, next week we can make it fooscheme, its
>> just the linking to the Big Graph (TM) that will be hard? Also since
>> renaming a taxonomy is an admin action, I don't think ALTER RENAME
>> would kill us here?
>
> There is no implied criticism or nitpicking in my question. I simply
> do not understand how the thing will work, and want to fix that lack
> of understanding.

I got that, its just that I thought the specific point you raised was
valid but easy to route around :-)

> I'll try to rephrase.
>
> The proposed scheme involves generating a schema which has an X_entry
> table, and one table per taxonomy. Are we calling all of the following
> things taxonomies:
>
> 1) COFOG
> 2) entities (in our current sense of transactional endpoint / graph node)
> 3) classification schemes used only a single dataset (e.g., "AccountGroup")
>
> and would they all give rise to table in your proposal?

Yes & yes. (Note we're talking about taxonomy names, not dimension names.)

- Friedrich