Re: control combination was: Re: protocol-22 comments)

[Hallvard B Furuseth <h.b.furuseth@usit.uio.no>]

Jim Sermersheim writes:
> So it seems there are three cases for a combination of controls:
> 1) the combination is documented in a specification
> 2) the combination is not documented, but is defined to act in a
> certain way in an implementation

What does (2) mean?  Someting like "the control specs do not document
the combination, but the implementation does"?

Will - and should - (2) cover the ManageDsaIT control from my later
post, and non-RFC controls like I mentioned?

> 3) is neither 1 or 2.
> 
> For #3, I think we are saying an error is returned.
>
> For #2, I think we need to state that behavior is undefined. This is
> because the lack of a specification can cause two implementations to act
> differently.

Yes, for (2) I think so... but I'll wait to hear what (2) means:-)
Similarly, I wonder what "not defined (or not known)" below means.

> "When a combination of controls is encountered whose semantics are not
> defined (or not known), the operation SHOULD fail with protocolError,
> otherwise the results are undefined."

If (2) is meant to be as liberal as my rewording above, we could add
something to the effect that operations also SHOULD fail if the
semantics of the combination of controls are ambiguous and not defined
in the control specs.  Or if you prefer, that implementations SHOULD NOT
define semantics for such combinations, which amounts to the same thing.

-- 
Hallvard