Lambda the Ultimate

Agreed on all points. This truly is a sad day.

Thanks for the link to the book on Bigraphs.

Sort of Tantalizing, almost, almost, what I'm looking for.

Here is my problem, perhaps you can advise me whether Bigraphs apply, or whether something else would be more suitable.

I have 2000+ C files which get compiled and linked into N products.

Fine grained managing and selecting that many files gets way too complex, so I have divided them into directories or "managable chunks".

These directories or chunks are the basic "units of assignment to a deliverable". ie. I can select a directory full of files and include it in (or exclude it from) any product.

Now there is a strict digraph between these chunks.

If a file x in set X refers to a symbol declared in file y in chunk Y, then X depends on Y. ie. If X is assigned to a product, so must Y.

Now all this works by far the best if...

* Everything is acyclic.
* Coupling is low.
* Cohesion is high.
* The definition of the symbol is in the same chunk as the declaration (or a weak symbol is provided).

I feel certain that the current assignment of files to directories is far from optimal, and that I could devise an algorithm to improve it.

Reading the excerpt from that book makes me wonder if the author hadn't worked on a similar problem. Certainly my problem is a constrained subset of all possible bigraphs.

Here is my problem, perhaps you can advise me whether Bigraphs apply, or whether something else would be more suitable.

While the description of your problem is such that I can imagine going about modeling it with bigraphs, I think they are probably several levels too abstract to give you a direct help in the short term.

(I seem to recall another thread here recently about dependency management; that would probably give you more directly helpful pointers.)

Bigraphs essentially take the pi calculus to the next level to "ubiquitous computation", and seem to have been taken up enthusiastically by the bio-computing crowd.

To be honest, my inner jury is still out on whether they are "the thing" or not, but there is no doubt they represent some brilliantly elucidated work.