Refine requirements on If-Unmodified-Since

[mnot]

Fixes #326

[reschke]

This is a normative change that we should discuss on the mailing list. It would be good to understand why this change is being made, and if we do it, we need to list it in the Changes section.

[mnot]

After merging this, we should align IMS and IUS with the approach taken in #357

[reschke]

That's a new normative requirement, right? I do not object, but it would need to be called out in changes-from-rfc723*.

[royfielding]

Perhaps it would be easier to state that the condition evaluates to false if the target resource has no selected representation or if the selected representation has no last modification date available, since both would indicate that the resource has changed.

[mnot]

It's already implied by the new clause above, I think it's OK to just delete this new requirement.

[royfielding]

Perhaps it would be easier to state that the condition evaluates to false if the target resource has no selected representation or if the selected representation has no last modification date available, since both would indicate that the resource has changed.

[reschke]

I still believe we need to call this out as normative change, even if it's almost just a bug fix.

[mnot]

Personally, I think that reduces the utility of the "changes from..." sections -- if we start noting changes this small, they're going to be swamped. I'm not going to refuse to put that in, but I don't think it's the right thing to do. @royfielding thoughts?

[reschke]

I consider the "changes from" sections the place where an implementor would need to go to see what might need to be changed in an implementation. (so yes, this is an edge case, but shouldn't we include it in this case?)