Re: [cap-dev] Implementation Questions

Deanna -

At 9:33 AM -0600 8/20/04, Deanna wrote:
>	Alert.sender is required to be globally unique.  However,
>alert.identifier does not have the same requirement.  Was this
>intentional?  We intend to base both the identifier and sender on our
>client's domain name.  Will this be a problem?

The idea was that each sender would ensure that the identifiers it 
used would be unique within its own domain, so that a concatenation 
of the identifier and the sender would form a globally-unique message 
id.  Seems like what you propose would be perfectly legal, although 
potentially a bit redundant.

>	Alert.scope:  What is meant by "known operational requirement?"
>Our system identifies members by organizational affiliation, such as
>Fire Department, and role, such as Battalion Chief.  Using these
>identifiers, along with location, we can create target groups.  Would
>this be an example of a known operational requirement?

Could be.  The point of scope is to flag those messages which ought 
not be disseminated beyond their intended audience.  "Restricted" 
means the desired limitation can be expressed as a rule (in the 
restriction element) while "Private" means it's expressed as an 
explicit set of addresses (in the addresses element).

So the point is really whether the information is in any way 
sensitive... understanding that there's no actual guarantee that 
message recipients will honor such dissemination-control assertions 
anyway.  Enforcing such restrictions would require a global trust 
architecture that doesn't yet exist.

Anyway, I'm not persuaded that this is necessarily the best way to 
address this issue... it's just the best we could come up with on the 
first go-round.  Mainly we wanted to help folks be explicit about 
their expectations as to whether any individual message was public or 
for-official-use-only.

>	Alert.scope:  It seems to me that addressing will be an obstacle
>that must be overcome.  I have seen the EPAD specifications, but it
>could be awhile before this is in place.  Has there been any discussion
>on specifying the use of a directory service, such as LDAP?

It's a dilemma.  On the one hand, as you point out, addressing isn't 
going to implement itself.  On the other hand, at the time we wrote 
the spec, and even now, there's no one universal dissemination 
network and no one universally-applicable addressing regime. 
(Although we did lean a bit, implicitly, toward the addressing style 
used in email and XMPP, among others.)

Basically, we tried to support whatever got implemented without 
lashing users to any particular platform.  Hopefully now that the 
issue is salient it will be addressed in one or more ways that 
everyone can live with.

- Art