At 12:10 PM 9/10/2001 -0700, Michael W. Condry wrote:
>Dan-
>We are looking more into service discovery for OPES, hopefully
>RUP will help. How do UpNP and RUP align?
you mean plug and play? Well, take WCIP, a cache could p-a-p if the web
server includes "Invalidated-By" in the HTTP response. Otherwise, a CDN
infrastructure has to exist, which will manage all the caches, including
bootstrap them with the WCIP session URLs.
Speaking of OPES + RUP on this regard, here is what I think.
RUP plans to associate (optionally) a "Location" directive with the objects
covered by the session. This "Location" may indicate the location(s) of
service(s) to those objects... so, it can be
"OPES-protocol-x://10.1.1.1:77/x/y/z". It may also be
"multicast-pgm://10.1.1.1:99/x/y/z". -- as long as the receiving cache has
the protocol module named "OPES-protocol-x" and "multicast-pgm", respectively.
Is this what you are looking for?
Dan
>At 10:10 AM 9/10/2001, you wrote:
>>Thanks, Michael!
>>
>>I suppose hand waving is one major cause of this type of problems.
>>Hopefully the RUP doc is dealing only with the "requirement" that there
>>needs to be a way for discovering sessions --- while the actual protocol
>>design will nail down the details, e.g., the Invalidated-By header in
>>HTTP response, as well as manual or out-of-band config of the resource
>>group's URL.
>>
>>At 08:52 AM 9/10/2001 -0700, Michael W. Condry wrote:
>>>Just in case you have not seen this.....maybe you
>>>should look at the discovery war that
>>>Scott is going to start...
>>>>Date: Sun, 9 Sep 2001 18:08:54 -0400
>>>>From: Scott Brim <swb@employees.org>
>>>>To: midcom@ietf.org
>>>>Subject: Re: [midcom] Topology considerations documents
>>>>Mail-Followup-To: Scott Brim <swb@employees.org>, midcom@ietf.org
>>>>User-Agent: Mutt/1.2.5i
>>>>Sender: midcom-admin@ietf.org
>>>>X-Mailman-Version: 1.0
>>>>List-Id: <midcom.ietf.org>
>>>>X-BeenThere: midcom@ietf.org
>>>>
>>>>On Thu, Sep 06, 2001 02:36:27PM -0400, Bob Penfield allegedly wrote:
>>>> > How the midcom agents learn the realm names is not a midcom problem. It
>>>> > could be through static configuration or some discovery protocol
>>>> between the
>>>> > entities sharing that realm naming scheme and namespace.
>>>>
>>>>This sounds a lot like economics discussions which begin with "First, we
>>>>assume a perfect market" :-). Think about what you're assuming!
>>>>
>>>>Static configuration? See what I wrote about the management headaches
>>>>and the possible security holes (which you wouldn't know about until
>>>>after a violation occurred). Imagine an ISP with thousands of edge
>>>>boxes that it needs to manage through these changes.
>>>>
>>>>Discovery protocol? For that to work, everyone would need to have
>>>>something discoverable. You would need the world's only globally unique
>>>>namespace, would require even casual users to get a name in order to
>>>>make a call, and then you would need the discovery protocol itself -- or
>>>>you would have to change every application signaling protocol in the
>>>>Internet. More arguments about this later.
>>>>
>>>>..Scott
>>>>
>>>>_______________________________________________
>>>>midcom mailing list
>>>>midcom@ietf.org
>>>>http://www1.ietf.org/mailman/listinfo/midcom
>>>
>>>Michael W. Condry
>>>Director, Network Edge Technology
>>
>>Michael W. Condry
>>Director, Network Edge Technology
>
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:22:59 MST