valueTesting optional and error cases

A while ago, I wrote about...

 specifying extended valueTesting without reference to implementations

but I eventually let it go.

Bjoern Hoehrmann commented on it in...

so let's take another look.

I can think of a few cases:
  - user-defined functions/operators
  - calling functions with args outside the domain

I still think we should make no reference to "implementations" in the QL
spec; just specify the semantics of the language. I'd like to keep
software conformance issues out of the QL spec and in the protocol spec.

Dan Connolly, W3C
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Sunday, 24 July 2005 00:45:34 UTC