Re:What Einstein Figured Out About Software Architecture
by fiddlemath on Thursday November 29, 2012 @06:09PM
(#4850861)
I understand and endorse the idea here.
It seems, though, that there must be some important distinction between "action at a distance" in physics, and "action at a distance" in software, that's being passed up.
In particular, "distance" in software is surely a function of "distance in mental space", rather than simply the number of lexical-scope boundaries crossed, or the specific number of containing modules. Do we have some way of formalizing this intuition of "distance", such that we could actually use it as a design measure?
Re:What Einstein Figured Out About Software Architecture
by mquander on Thursday November 29, 2012 @06:35PM
(#4850957)
In particular, "distance" in software is surely a function of "distance in mental space", rather than simply the number of lexical-scope boundaries crossed, or the specific number of containing modules. Do we have some way of formalizing this intuition of "distance", such that we could actually use it as a design measure?
I think those metrics should converge on each other as you refine the shape of your program toward the Right Thing.
Rephrased: Your module or scoping system serves you well to the extent that you can make hierarchies of modules and scope correspond to hierarchies in your mental model.
Re:What Einstein Figured Out About Software Architecture
by fiddlemath on Thursday November 29, 2012 @06:13PM
(#4850883)
er, yes. "Locality" here doesn't mean keeping everything in the same place, physically - it's talking about keeping causes in the same modules as their effects. Different "space" entirely.
It seems, though, that there must be some important distinction between "action at a distance" in physics, and "action at a distance" in software, that's being passed up.
In particular, "distance" in software is surely a function of "distance in mental space", rather than simply the number of lexical-scope boundaries crossed, or the specific number of containing modules. Do we have some way of formalizing this intuition of "distance", such that we could actually use it as a design measure?
I think those metrics should converge on each other as you refine the shape of your program toward the Right Thing.
Rephrased: Your module or scoping system serves you well to the extent that you can make hierarchies of modules and scope correspond to hierarchies in your mental model.