Re: [Chicken-users] s11n and tinyclos

[Kon Lovett]

On Mar 7, 2006, at 10:27 PM, felix winkelmann wrote:

> On 3/2/06, Simon St-Onge <address@hidden> wrote:
> > Hi,
> >
> >  I am trying to serialize/deserialize a tinyclos object using s11n:
>
> Hi, Simon!
>
> Well, looking at the dumped instance with chicken-dump appears
> to show something sensible, but the problem lies deeper: TinyCLOS
> instances contain direct references to their class, which gets  
> serialized
> with the intance itself. But on deserialization, the class loses  
> its identity,
> and therefore the whole class-hierarchy of a deserialized instance
> is invalid (we have several <class> or <top> classes in the end).
> TinyCLOS has no notion of a global class table, which could be used
> to map deserialized classes back to their "canonical" (i.e. system- 
> wide)
> identity.
> So the bottom line is (currently) that instances (and thus classes)  
> can not
> be reliably deserialized.

Or anything that has a global identity, beyond built-in. Need some  
kind of '(remain <object>)' to designate existing "root" objects,  
i.e. objects that will exist when the de-serialization occurs. How to  
handle the "root" dependency is unspecified.

> cheers,
> felix
>
>
> _______________________________________________
> Chicken-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/chicken-users