Re: Y Store /Closures

[Daniel Weinreb <address@hidden>]

Jacob Matthews wrote:

>This is actually just a problem of marshalling continuations. If you can
>arrange to be able to do that (hopefully space-efficiently!), you can just
>have your web server or CGI script marshall any continuation it'll need to
>keep around and stash it on the outgoing webpage or in a database
>somewhere. 
>
Yeah, it's the problem of marshalling them and then storing them and 
then retrieving them
and then unmarshalling them, space-efficiently and time-efficiently, 
ideally in such a way
that it works for any continuation in the language, and so that you 
really get back to the
state you were in when you started.  This can be hard for a lot of 
reasons since various
kinds of state are hard to marshall, e.g. if they depend on various 
kinds of runtime context
like already-open streams and so on.

>At that point 
>
(If you get to that point!)

>you're in the same world that traditional
>solutions are in, so crash-recovery/load-balancing/etc ought to work just
>as well. See Graunke et al., "Automatically Restructuring Programs for the
>Web" (http://www.ccs.neu.edu/scheme/pubs/ase2001-gfkf.ps.gz) which
>describes a technique, though the paper is written with a
>small-to-mid-size rather than an enterprise-level website in mind.
>
Yes, this work is very directly relevant to what we're talking about.

>  
>