I'd like to request that the LSC take up the topic of a connection standard. Also, since Sergio was first to the develop something like this, I think his way to doing things should probably be what is standardized.
And here is probably the first one that got implemented:
http://news.lugnet.com/cad/?n=9002
That said, since SR3D is clearly the first successful one, I agree that we have at least a de-facto standard to work from...
Personally I think defining all connection types and positions is 'old fashioned' (although still desirable). It might be more interesting to research needed meta data for implementing a physics engine approach, which could (potentially) do the same and more.
That's a personal dream though