RE: Wildcarding query

["Patrick Lidstone \(Personal e-mail\)" <patrick@xxxxxxxxxxxx>]

> Spec says:
>
> "Unless explicitly specified, the : (colon) sub-address
> separator is treated as being synonomous with a . (dot)
> separator. Exact matching against a : (colon) is only
> required in situations where wild card filtering against just
> the sub-address element is required."
>
> What does that last bit mean?  When does it apply??
>
> Given a device a.b.c:d.e
>
> a.b.*:*.* would shirley match.  Is this what "explicitly
> specified" means?

If there is no colon, then it still matches.

a.b.c.d would match a.b:c.d, a.b.c.d, a:b.c.d etc.

The reason for this is that the decision to group subaddresses into a
single logical device/address space is completely arbitrary and at the
discretion of the hardware designers.

> But does a.b.c.d.e match it?
>
> How about a.b.*.*.e   or  a.b.c.*.*  or a.b.*.d.*

Yes to all the above.

> I'm also a bit hazy on body wilcarding.  In the BSC spec it
> is clear that multi-part bodies are permitted (xAPBSC.cmd
> example), and the xAP spec says "Any multi-block message
> [note message not schema] with a wildcarded header address
> may optionally include additional target filter" but the BSC
> spec makes no mention of body targets.
>
> Is it safe to just generically assume that _any_ body may
> contain a target= element?

There was talk about deprecating this at one point, but AIUI the spec
currently allows for it. Kevin?

> Finally, the '>' character, it is correct that nothing can
> follow it in a wildcard specification?

Only if there is no subaddressing, I think.

a.b.c.>.d would be ILLEGAL.

But

a.b.c.>:d would be LEGAL.

At least as I have implemented it.

I have to say, the wildcarding schemes - whilst comprehensive - are
perhaps over complex, and might benefit from a bit of rationalisation.

Patrick