> Jay Bryant wrote:
> If a language has a feature, the specification should state how that feature works...let's write a spec that doesn't call for anything to be implementation-dependent.
For the record, I agree. If I gave a different impression, I may have made my point hastily.
I do think there's room, though, for extension here. Users have very diverse needs when it comes to xrefs and we need to be able to accommodate those scenarios. We already have an extension point at the include processor (putting aside for a second that it needs improvement). This allows implementations to create an augmented contract with users to resolve includes from locations that are not dependent on the local file structure. I think we could have exactly the same thing for the xref macro. But it would be clear that doing so is an extension and the user is knowingly entering into that augmented contract (and perhaps we can even standardize that mechanism). The standard behavior would still need to be clearly specified. I'm hoping that fits within the bounds of what you find acceptable for a standard.
Best Regards,
-Dan
