Re: draft-ietf-impp-pres-01

On Wed, 2003-01-15 at 16:34, Dave Crocker wrote:
> Robert,
> 
> Wednesday, January 15, 2003, 12:12:18 PM, you wrote:
> Robert> I mean case 2.
> Robert> Comparing URIs is seldom as simple as comparing bit-streams.
> 
> When they are used as unique identifiers (URN), rather than semantic
> addresses (URL), why are they not just bit strings?

At risk of jumping to fast to the concrete:
_IF_ we scoped SubscripID to watcher,target, would you expect
(1,pres:[email protected],pres:[email protected]) to match
(1,pres:[email protected],pres:[email protected])?

> 
> 
> Robert> As I think you are saying, if you scope the identifier to a context
> Robert> then the cost of recognizing that identifier includes the cost of
> Robert> recognizing the context.
> 
> right.
> 
> Robert>  However, if recognizing the context is more complex
> Robert> (comparing URIs for example) we have something real to consider in the
> Robert> tradeoff against the properties Thanos was asking for.
> 
> URIs are self-defining.  What "context" is there to recognize, and what
> do you mean by recognize?
By recognize I meant do the computation necessary to identify the 
context. In the concrete case we're dancing around, it means finding
and comparing pres: URIs for the watcher and target (not against each
other, but against the watcher, and target URIs for the set of 
subscriptions I currently know something about).

> 
> 
> Robert> For those cases where identifying the context is simple, I would
> Robert> prefer to avoid small increment in interop risk due to having to
> Robert> look at multiple things in order to recognize the identifier.
> 
> relying on context means that you must look at multiple things.  one is
> ensuring that you are interpreting in the right context.  the other is
> the identifier.
> 
> 
> Robert> For those cases where identifying the context is more complex, I
> Robert> anticipate that the evaluation of effort to build a self-defining ID
> Robert> vs effort to identify context will almost always fall out in favor
> Robert> of building a self-defining ID.
> 
> we agree.
> 
> 
> Robert> If I understood him, the specific context Thanos was proposing was
> Robert> scoping the SubscriptID to To and From (I am supposing he means
> Robert> watcher and target from impp-pres-01). Comparing these fields 
> Robert> involves comparing pres: URIs. Comparing pres: URIs is much harder
> Robert> than comparing bit streams.
> 
> I think folks need to construct some concrete examples, because I do not
> see the truth in what you are asserting.
See the first question above.
> 
> 
> Interactions between watchers and servers take place over multiple
> connections and potential over a long period of time.  This calls for
> self-defined identifiers.

We're arguing hard for the same conclusion :)

> 
> d/
> -- 
>  Dave <mailto:[email protected]>
>  Brandenburg InternetWorking <http://www.brandenburg.com>
>  t +1.408.246.8253; f +1.408.850.1850
> 
> 
> 
> 
>   [reminder: [email protected] for non-technical discussions, please]





  [reminder: [email protected] for non-technical discussions, please]