[cap-talk] Selling capabilities programming

[Jonathan S. Shapiro]

On Wed, 2007-07-25 at 22:40 -0700, David Wagner wrote:
> While user sessions might last hours or days, process lifetimes are
> often significantly less than that -- often enough that I think it might
> potentially still provide some useful (albeit partial) protection.
>
> What do you think?  Are you persuaded, or do you think I'm missing
> some important points?

I think it depends on a lot of factors that are hard to predict. I
frequently have emacs sessions that run for a week or more. In some
software, closing the last document window does not cause the
application to exit -- the application delays its exit, seeking to avoid
startup costs if you open another document soon.

I suppose I think that it is not a good design practice to depend on
exit behavior. Even if the dependency is "soft", it becomes part of the
designer's tacit assumptions, whereupon it will get violated by a change
in the application's behavior.


shap