RE: DR 203: rewording?

> I tend to agree with dropping DR203. Perhaps the intent of this requirement
> was to
> ensure that the RPC convention did not utilize semantics dependent on a
> particular
> language or object system? If so, I think DR 201 covers this sufficiently by
> requiring
> straightforward mappings to languages and object systems. 

I disagree that this is covered in 201.  Straitforward in no way implies 
interoperable.  Creating a single binding will be trivial if the same 
implementer and model supplied the encoder and decoder.  Creating a 
second, which has to try to interoperate with the first, may be a bigger 
challenge, and a third may be more problem, still.

I can just see, for example, Microsoft arguing with IBM or Netscape over 
the way their particular RPC gateway bound to the wire protocol, and who 
is right.  Creating a wire protocol that is not designed for 
interoperable bindings that can be tested is wrong, IMO.

Ray Whitmer
rayw@netscape.com

Received on Wednesday, 15 November 2000 19:10:44 UTC