Re: Interoperability for MUST, SHOULD, MAY

[Randy Bush <randy@psg.com>]

> Not sure if we have reached (rough) IESG consensus on this.
> I think we do have consensus on:
> 
>    We need to see two genetically independent implementations 
>    for every MUST and every SHOULD. 
> 
> Not sure we have consensus onL:
>    Harald added that potentially it even means we need to
>    see two genetically independent implementations of
>    ervery MAY as well.

specs have features.  the only way we have of testing the correctness
of specs is to test interop of the features.  as we are publishing the
entire spec, we need to test ALL the features.

randy