That is a good quantitative question. The 'sling less code' is the metric I was using.
I would love to come up with a good architectural quality test that was more quantitative. I've got a script that can go count how many times a given function or module was rewritten but running that on our code base of a few million lines of code really doesn't correlate with a 'gut' feeling of good architecture or bad architecture. For example, the Blekko crawler has essentially been rewritten from first principles three times over the 7 years we've been doing this. Generally that was not a result of "bad architecture" so much as it was unknown unknowns, basically writing a crawler that can run unattended for weeks or months is a really complex and often subtly nuanced problem. And sometimes is clear that the rewrite was just because someone didn't understand how something originally worked and so re-implemented it in a more understandable way. What I'm getting to is that code "touches" are not the quantitative measure I'm looking for.
I would love to come up with a good architectural quality test that was more quantitative. I've got a script that can go count how many times a given function or module was rewritten but running that on our code base of a few million lines of code really doesn't correlate with a 'gut' feeling of good architecture or bad architecture. For example, the Blekko crawler has essentially been rewritten from first principles three times over the 7 years we've been doing this. Generally that was not a result of "bad architecture" so much as it was unknown unknowns, basically writing a crawler that can run unattended for weeks or months is a really complex and often subtly nuanced problem. And sometimes is clear that the rewrite was just because someone didn't understand how something originally worked and so re-implemented it in a more understandable way. What I'm getting to is that code "touches" are not the quantitative measure I'm looking for.