Re: RFC 47 (v1) Universal Asynchronous I/O

John Tobey <[email protected]> writes:

> On Tue, Aug 08, 2000 at 10:39:23AM +0200, Bart Lateur wrote:
> > On Mon, 7 Aug 2000 21:45:28 -0400, John Tobey wrote:
> > 
> > >But the above code would be possible if get_submission stored the
> > >current continuation in a database before returning to the web server
> > >loop.  The page handler could associate the next request with the
> > >continuation and simply jump back into it when the form was submitted.
> > 
> > Ooh, not only must you have continuations, but you have to be able to
> > store it in a database as well? That implies that not only you must be
> > able to stringify the state (stack+variables), but to restore it from a
> > string as well. Something along the lines of the mechanism behind
> > AUTOLOAD?
> 
> The database is only needed if another process is expected to handle
> the next request.  Otherwise, you can store the continuation in a
> variable and include live filehandles, db cursors, etc.

Given that a continuation is going to store the state of the
interpreter doesn't a serialized continuation have the potential to
get very large indeed? Which may well be a bad thing.

-- 
Piers