All, as you know if you are on the squeak-dev list, the list server is 
still choking on bad DNS records  (nsa hacked?) and there are some lost 
messages....actually, do you have to be a member of the squeak list in 
order to post there, too?   there has got to be a better way to 
cooperate!  ;-)     so anyway, this is why I haven't posted more to 
respond to the great suggestions.

David and Tyler, because of this, I will respond to your emails 
tomorrow, but I quickly added some thoughts after I respond to Avi's 
note.

On Sunday, January 19, 2003, at 08:04 PM, Avi Bryant wrote:

>
> On Sun, 19 Jan 2003, Robert Withers wrote:
>
>> hmm...doesn't this use SmartRefStream?  The problem is that E uses 
>> Java
>> serialization and we don't do that - we have Squeak serialization.
>
> Why not implement Java serialization in Squeak?  It's been done in 
> Lisp,
> for example (http://www.pointnclick.com/lijos/), and might lead to some
> other useful integration (Java RMI in Squeak?).
>

Avi,  I think that's a good idea too.  I tried Gorans java scanner, but 
it had nothing to do with serialization.  What would be really nice 
would be a completely pluggable encoder stream.  You could plug-in type 
encoders for primitives and constructed types.    <craziness>You may be 
able to exchange code too.</craziness>  You could write support for 
java, WOS, CDR, SIF, Squeak, ANS.1

My concern is that I am busy in other parts of SqueakE, and I am just 
going to run with ReferenceStream for the time-being.  If we decide 
that a change is appropriate, then both us and e would have to change 
protocols.   As such, this thread represents my attempt to make people 
aware of where things are.  It's a forum to strategically and carefully 
decide what the best way to go about this is.  Also, many people have a 
stake in this decision, so it should be made with their input.

I have a preternatural hope that some few may decide that this 
represents a very interesting possibility and pick up the challenge of 
some difficult work and post a CDR or Java or WOS encoder.  :)



Tyler, A comparison matrix sounds like just the thing we should do.   I 
really, really like your idea of a compatibility graph.  There is every 
reason we could negotiate these over the wire, in peer2peer fashion.  
Just as the naming service has been embedded, so could the interface 
repository.  Regarding your implication that a signature has to define 
what exceptions it can throw, and i know it was just an example, but 
assume that all exceptions are runtime exceptions in squeak.  We don't 
declare that information in the signature, because the signature could 
be implemented by anyone and they may throw different exceptions.  :)


David,  Totally!   :)   thanks very much for your differentiation of 
the message format from the serialization format.  Do you have a link 
to CapIDL?   Does anyone have one of those nifty open source IDL 
compilers in squeak?!  ;-)


regards,
Robert