🚀 go-pugleaf

Has anybody thought about suitable distance/similarity/dis-similarity
measures for reverse engineering large s/w systems.

I am thinking of taking inter-module reference information,
building a distance matrix then using clustering and/or
2D projection techniques (eg metric scaling).

The question is: how do you produce a distance matrix that
will reveal a *layered* software architecture [if it exists]?
Presumably the modules in a layer may well have little 
inter-communication; the commonality is, presumably, in their
callers and who they call.

Any ideas?

Regards, Peter (P.G.Hamer@bnr.co.uk)