Scott Ferguson skrev:
> It needs to be possible for a 2.0 server to understand 1.0, i.e. the  
> 2.0 spec needs to interpret 1.0 requests as 1.0.  I don't know if the  
> backwards compat should be "must" or "should" for a server  
> implementation.

Actually, I'm just checking for the version number. If that is not 
exactly 2.0 I'll return an error. I'm not aiming at all towards 
supporting any 1.0 "clients". On the other hand, implementing a 1.0 
encoder/decoder is under 500LOC, so why not .. hmm...
We have an implementation for 2.0 and it's hardly over 2000LOC's yet, 
and it's fairly complete. It's was well worth the time spent writing it. 
It is extremely much faster and more compact for our kind of data-heavy 
communication (we send well over 100K objects in some requests). I'll 
try to write up a "case study" in an article if I get some time over.

Anyway, the point is that even if a client connects with 1.0 to this 2.0 
server, it will not understand the response anyway, because that is all 2.0.

> After our 3.2.0 release, I'll go  
> through the spec once again, ask for feedback here, and then submit it.

Sounds like a good plan. I haven't checked the source for 3.2.0. Is it 
based on 3.1.*? That case, I'll might just go forward and test against 
the snapshots instead. Perhaps implementation feedback would be better 
for you to receive on that version?
Will you do any more releases of the 3.1 branch?

"Scrim" will be released this week as a fully working "alpha". The only 
reason why it is alpha is because we might change the client-proxy-stuff 
and add a little bit more SDK stuff for the serializer/deserializer 
stuff. The protocol implementation itself will be in production starting 
from Monday.

Cheers,
/Niclas


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

Reply via email to