Re: [cap-dev] Question re: alert.sent

[Art Botterell <acb@incident.com>]

Good question, Deanna.  Let me take a whack at it.

The message-wide <sent> element is the only required timestamp on a 
CAP message... it should refer to when the message was actually 
approved and issued by the originating agency.  In that sense the CAP 
message format is opaque as regards any internal approval processes 
(but see below.)

If none of the other time-based elements are present, then the <sent> 
time is also, by default, the time the message should be delivered to 
its final audience (in your example, posted to the Web).  However, 
sometimes it may be desirable (again, as in your example) to embargo 
dissemination till later... or even for multiple <info> blocks within 
a single <alert> message to have different release times.  In that 
case the <effective> element within <info> can be used to provide a 
"start time" value.

It may also be desirable to indicate the forecast time of occurrence 
of whatever is being described, even when the message is released 
(effective) earlier... e.g., a forecast arrival of a storm.  So we 
also have the <onset> element.

Seems like how your particular application would map to all this 
would depend, in part, on the relationship between the "creator" and 
"approver" roles.  In a tightly-coupled environment... e.g., inside a 
single agency... you might not want to finalize the CAP message until 
the approval is given, in which case <sent> would be the approval 
date.  In a more loosely-coupled... e.g., multi-agency... process, 
there might even be two CAP messages... an initial "private" one (as 
defined in the <scope> element) which triggers a second "public" one 
issued by the approver.  Much depends on the precise meaning of 
"approval" in a particular process... does it comprise validation of 
the information or just an editorial decision to make the information 
public?

Also, it seems like the "notified" date/time you mention wouldn't 
really be a component of the original CAP message, since its values 
would vary from recipient to recipient depending on system 
performance... and in any event they wouldn't be known until after 
the message had been composed and released.  However, where there's a 
requirement to track individual or aggregate delivery events, those 
might be transmitted in reply CAP messages of <msgType> "ack" with a 
<references> pointer back to the original message, and the 
system-specific results in <parameter> tags.

At least, that's how it strikes me. ;-)  HTH,

- Art


At 3:22 PM -0600 8/31/04, Deanna Rider wrote:
>I have a question about the sent element.  I am trying to interpret the
>meaning of "origination" of the message based on our process:
>
>*	A user creates an alert and submits it to another user for
>approval.  This action generates a creation date/time.
>*	A user approves an alert for publication and notification.  This
>action generates a posted date/time.
>*	When the alert is approved, the approver assigns a start
>date/time to an alert that determines when it should be published on a
>website and when notification of the publication should be sent to
>designated recipients.  This action generates a start date/time.
>*	Notifications (e-mail, phone, pager, etc) are sent to the user
>when the start date/time has passed.  This action generates a notified
>date/time.
>
>I am guessing that the date you want is the posted date, but I hate to
>take actions based on assumptions.  Thanks in advance for your
>assistance.
>
>Deanna Rider
>Product Development Director
>www.invizeon.com