[Concordia] Organizing use cases

[Eric Norman]

On Jun 18, 2007, at 5:33 PM, Eve L. Maler wrote:

> Eric Norman wrote:
>> Here's one suggestion that might help organize it better.  Each
>> row represents a protocol that needs to make adjustments to 
>> communicate
>> with the protocols along the columns.  The columns represent protocols
>> that can be used as is.
>> For example, look at the OpenID/SAML case.  The first two clauses 
>> (OpenID
>> makes SAML assertions and uses SAML authentication context would 
>> belong
>> in row OpenID, column SAML (where they are now).  The third clause 
>> seems
>> to call for an adjustment to SAML deployments, hence would be in row 
>> SAML,
>> column OpenID.
>
> Y'know, I honestly hadn't gotten as far as realizing that I didn't 
> really need one whole triangle of this thing. That's sad. :-) 
> (Luckily, I'm on vacation and offline all next week!  Hopefully that 
> will cure me.)
>
> Your suggestion is very interesting.  I'm a little worried, though, 
> that it will pre-judge technical solutions.  I had intended to list 
> them without consideration of "who has to change", especially when we 
> don't have a lot of detail on the use cases yet.

I don't see it as pre-judging at all.  There would be cases where the
adjustments could be made at either end.  I.e. OpenID speaks SAML or
SAML speaks OpenID.  In such cases, there would be an entry in both
row/column boxes on opposite sides of the diagonal.  The fact is
that adjustments are going to have to be made somewhere; I'm just
trying to help capture what they are and where they could be made.

> For example, "SAML attribute assertions in OpenID" could involve a 
> combo effort: someone doing an OpenID extension that allows for SAML 
> content to come along for the ride, and someone doing a SAML attribute 
> profile intended for use with the extension.  Or maybe the whole thing 
> gets rolled together in a single spec, done in the IETF.  Should that 
> go in horiz OpenID/vert SAML, or vert OpenID/horiz SAML, or a third 
> dimension entirely?

In the "combo" case.  The combination could be "factored" with each
factor going in the appropriate box.  What this would not capture
is that these two factors are linked.  Color coding?  Footnotes?
Dotted lines across the diagonal?

Eric Norman
http://ejnorman.blogspot.com