Dnia 16-12-2007, N o godzinie 03:05 -0800, Paul Rubin pisze:
> > Marshalling forces evaluation of the values marshalled.
>
> I don't see why any fundamental reason this is necessary. Nothing
> wrong with writing out unevaluated thunks in some fa****on.
Trying to write out an unevaluated thunk, even if we somehow solve
serializing program code, has a fundamental problem of deciding what
to consider a part of the given value.
If the code of evaluating a given integer uses a function, is the code
of this function a part of the integer? Should we emit the code which
implements integer addition, if the integer uses addition?
Depending on what to choose, the value might readable only by the same
exact program, or only by programs using the same libraries, perhaps
only running on the same architecture... And the serialized value might
be huge, even if we manage to share common parts of separately written
values.
I don't believe in serializing unevaluted thunks at all.
--
__("< Marcin Kowalczyk
\__/ qrczak@[EMAIL PROTECTED]
^^ http://qrnik.knm.org.pl/~qrczak/
|
