[plt-scheme] Continuation marks question

Or maybe if you have sufficiently powerful inspector.

Robby

On 5/12/07, John Clements <clements at brinckerhoff.org> wrote:
>
> On May 12, 2007, at 11:00 AM, Jay McCarthy wrote:
>
> > (Really a message to Matthew, but to the list for posterity.)
> >
> > Is it possible to a get a function with continuation-marks-set->list
> > that does not require me to know the keys in the set?
> >
> > There are a few ways I can get around it if not. I can keep track of
> > all the keys I see in the expanded program, but that won't do for
> > marks put in my called modules. I might be able to work with just
> > exposing the parameterization key.
> >
> > In a sense this is a bad idea, because I don't know the key for a
> > reason. If that's the answer, then I would at-least say that the
> > parameterization key should be available.
> >
> > Jay
> >
> > ps What I am doing is serializing continuations... my current
> > implementation loses the continuation-marks, which manifests by users
> > not being able to use parameters, which I think are very useful on the
> > web.
>
> Yep, this was part of the design of cm's.  That is, if you want to
> capture the user's cm's, you need to know what they are.
>
> One sneaky trick you could play with separate modules would be to
> redefine with-continuation-mark.  Of course, this only works if you
> have access to these modules' source.  Alternatively, you could
> insist that in order to be serializable, these modules include a "cm-
> keys" call that is guaranteed to return the set of all keys used by
> the module. Or or or...
>
> WRT the parameterization key: perhaps it would be possible to add to
> mzscheme a "with-parameterization-key" form that uses the
> specified ...  hmm, that would pretty much destroy the security of
> parameters used by called modules.
>
> Well, Matthew will probably have something more helpful to say, but
> it appears to me that in general, the power to observe the full
> continuation comes only with access to all the source code, and
> that's probably as it should be.
>
> John
>
> _________________________________________________
>   For list-related administrative tasks:
>   http://list.cs.brown.edu/mailman/listinfo/plt-scheme
>