Re: TMDM and merging associations

On Tue, Sep 07, 2010 at 01:01:18PM +0200, Lars Heuer wrote:

> I agree that reification is underrated but I don't see a real use
> case for reifying roles because I think that associations should be
> immutable entities and that the roles should be absorbed by them.

I cannot agree more. (Well I could, but then I would have to kiss
you. Brrr ;-)

> Given that an association is an immutable entity, it makes no sense to
> reify a role to state (for example) that the particular role was added
> at date X since that date would be equal to the date when the
> association was created / added.

Generally, a reification such as

   "P plays the role FATHER"

is meaningless as it lacks to reify the context (other players, etc.).
With merging such a reification would possibly end up in a different
context (more players) than the original author has intended. 

I think the mistake made here is to view TMDM as a navigation
model. But _reification_ is a semantic notion, not just a link.

ad Steve P: Reifying roles makes "expressiveness" worse, not better. I
might want to add that the original idea of having (contextsensitive)
ROLES is still an excellent one. But they are underrepresented in
TMDM. Especially if you contrast this with "roles/traits/interfaces"
in modern programming languages. With TempleScript (which, btw,
understands roles) I had to fix the model too.

> Nice side effect: Several models outside TMDM would work more easily,
> i.e. Lars Marius' Q5 or the TMRM -> TMDM mapping.

Sigh, good times :-)

\rho