FW: [CAP] Unique Message Identifiers in CAP

["Bob Wyman" <bob@wyman.us>]

    The CAP specificiation, while saying that message identifiers are
supposed to be unique, doesn't say anything explicit about the scope
of the uniqueness. Normally, specifications for Internet protocols
make this explicit to avoid confusion and diverging implementations.
Is it expected that a message identifier is globally unique? (i.e. no
two senders should ever generate the same message identifier.) Or, are
these identifiers only unique for each sender? If only unique for each
sender, then should we assume that that *actual* identifier is
constructed by concatenating the message identifier with the sender
identifier?

    In any system that relies on unique identifiers, it is important
to identify the time period during which uniqueness is guaranteed.
However, the CAP spec doesn't do this. Are senders allowed to reuse
identifiers? If so, under what circumstances? Is there, for instance,
an implied period of time during which the uniqueness of identifiers
should be maintained? (Note: NWS seems to reuse the same message
identifiers whenever they generate info for a particular area. Thus,
the message identifier "wwa_California" is reused in *every* CAP
message that they write concerning California. Is this what the
designers of CAP intended? (Note: I realize that NWS is still using
0.9a1, however, the general issue still exists.) Can/Should message
identifiers be reused? I would suggest that this should be strongly
discouraged. If message identifiers are constantly reused in the
manner that NWS does, then they aren't really "message identifiers" --
rather, they are serving the purpose of "subject," "topic," or
"series" codes.
    If it is important to be able to identify a sequence of unique
messages that all address the same subject then support for that
should be provided for in the specification rather than having people
overload a field which should be message-specific.

    Because CAP message ids are assigned by senders, there are a
number of opportunities for severe confusion. For instance, it appears
to be difficult to take CAP feeds from both the NWS and California
EDIS without getting very confused about what is really happening. The
difficulty is caused by the fact that California repackages NWS
weather alerts as CAP messages which have different message
identifiers than the NWS CAP feed uses. Ideally, it would be possible
for the EDIS messages to at least identify the NWS messages that were
their source so that software could filter out duplicates. (Note: The
current "source" field is inadequate for these purposes.) The current
situation would force anyone who was reading both EDIS and NWS CAP
feeds to simply ignore all data that came from EDIS since it is
probably a duplicate of NWS data. This means that if California ever
issued its own "MET" alerts, not simply copies of NWS data, readers
would probably ignore them. This is not good.

        bob wyman