Scott,

On 1 Mar 2008, at 17:09, Scott Ferguson wrote:
>
> I'd think you'd see the same benefits if the types were handled
> lazily, just by making the transmission scope explicit and extending
> it beyond the call/reply.

Agreed.

I wasn't actually necessarily proposing the eager sending of the type  
negotiation per se. I was envisaging some kind of idempotent strategy  
where a sender can transmit references to types that he hasn't  
necessarily already communicated. So for example, in the sender's  
initial call, he just sends a reference straight away. Obviously the  
receiver won't know what to do with it. The receiver will have to then  
query the sender for the type information pertaining to the reference  
he received - a remote hash map lookup back to the sender. This  
eliminates the need for any transmission context that the sender and  
receiver would have to be aware of. Neither peer would have to know  
when a transmission context starts or ends and the receiver could re- 
request type information at any time. You would probably have to think  
about the buffering capabilities this would require - when the sender  
can resend or whether the receiver can buffer data it can't decode  
until it receives the corresponding type information.

> Assuming we can come up with a simple extension, I'd think this scope
> could be added to the Hessian 2.0 spec itself.  It's about time for us
> to clean the spec up and submit it to the IETF.

That would be really good. As indicated in my post, I wanted to follow  
up the article with a more detailed look at the protocol details, so  
we can definitely get something going. Obviously I want to get your  
input to get the simplest solution possible.

Ben


_______________________________________________
hessian-interest mailing list
[email protected]
http://maillist.caucho.com/mailman/listinfo/hessian-interest

Reply via email to