[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Y Store /Closures




On Sat, 1 Mar 2003, Daniel Weinreb wrote:

> I've sort of been interested in this for a long time, since I was in the
> "persistent storage of programming language data" game from about 1985 -
> 1999 or something.  There are some programming langauge data constructs
> that are easier to "store persistently" than others, and things like
> continuations and coroutines seem to be high on the list. We can debate
> about "acceptable level of inconvenience" in this case, but in general
> the problem of saving state persisently seems to be one drawback to
> modelling state as continuations.

Yes, I think this is true.  A related problem has to do with garbage
collection - if you're not extremely careful, a continuation is apt to
hold onto a lot of objects that you would normally think of as transient,
so the memory footprint for a session is usually higher than it would be
with other techniques (it's possible this could be avoided with a smart
compiler, but the continuations I'm working with are purely at the library
level, and so the implementation is quite naive).

For me, the benefits of using continuations outweigh these costs, but
everyone needs to evaluate the tradeoffs for themselves.

Avi