Date: Thu, 25 Jan 2001 10:23:04 -0800 Reply-To: Tim Kindberg <firstname.lastname@example.org> From: Tim Kindberg <email@example.com> Subject: Re: NID registration for ISBN In-Reply-To: <20010125111609.F418@bailey.dscga.com> Content-Type: text/plain; charset="us-ascii"; format=flowed
At 11:16 AM 1/25/2001 -0500, Michael Mealling wrote: >On Thu, Jan 25, 2001 at 10:43:35AM -0500, Leslie Daigle wrote: > > As Michael mentioned, there has been interest in contextualized > > resolution as a separate work effort. That will provide the relevant > > forum for discussing much of what you're getting at. > >And that forum can be found at: >http://lists.research.netsol.com/pipermail/c15n/ > >And the subscription info is at: >http://lists.research.netsol.com/mailman/listinfo/c15n
Thanks for that links. We've been working on contextual resolution as part of our 'CoolTown' system (http://cooltown.hp.com/): we map identifiers found on physical entities to contextually relevant URLs.
This is a good opportunity for me to try to spell out what has been bemusing me about the URN/URI community, looked at from the point of view of our work. When I started to work on resolution I naturally looked at what the URN community has been doing, but I tripped over the first assumption I came across: that URNs are 'location independent identifiers'. That is, the goal seems to be a system that would give the same 'authoritative' (the URN community's word) answer to everyone, every time, wherever they are.
That's the opposite of what seems to us to be the makings of a useful system, given our assumptions. Here's an example to illustrate our thinking:
Suppose I, you (another shopper) and a supermarket employee all pick up identical cans of beans in the supermarket, and suppose we all scan the barcode on the can with a wirelessly connected device to look at web pages about the beans (this is something we routinely do in CoolTown).
We think that the chances are that those three users would want to see different results. I want to avoid genetically modified (GM) foods so I want to see appropriate links to check the beans' status. You suffer from diabetes, so you want to see diabetic links about the beans. The employee wants to do a Safeway price or stock check. Actually, we may all want to see a link to the Safeway page, in addition to the other specialised pages I've mentioned.
And if we were back in our homes instead of the supermarket, then we may want to see yet another set of pages when we scan the beans' barcode: say, a page enabling us to put that item on our web shopping list.
In CoolTown we have ways to pick up the local context (Safeway, the home); but users also configure their devices to use more generally applicable contexts (e.g. GM foods and diabetes).
One name -- let's call it upca:78996800002 -- has bindings in multiple naming contexts: in each case, to a record containing a URL (we're considering Xlink).
Which of the bindings is 'authoritative'? Our answer is: they are all equally authoritative. Bindings derive from organisations that assert them (Safeway stores, The Campaign against GM foods, the Finnish Diabetes society, Tim Kindberg's household). Those organisations can digitally sign them to make them literally authoritative.
What about the fact that the name contains 'upca' -- doesn't that signify some 'extra' authority for the manufacturer? Our answer is: No, not _binding_ authority. There's a basic distinction between the authority to mint names and the authority to bind them to resources, which rarely seems to be made. The UPC people have devised a way of ensuring that manufacturers can mint names for their products without fear of collision. Those manufacturers then bind those names (a) literally, physically as part of the fabric products (which probably would count as 'authoritative' in court) and (b) virtually, to virtual resources about the products.
Should that mean that The Campaign Against GM foods _cannot_ also bind the UPC code to their own virtual resources, or that, if they do so, their binding has a lesser status?
I can understand why Heinz might wish that virtual binding didn't exist or was deprecated, for commercial reasons. But there's no logical objection that they can raise.
At least, not as long as we satisfy the requirement that no-one will get confused about whose binding is whose. And that problem can be solved by making bindings first-class, standard, digitally signed objects.
I sometimes wonder whether having UPC codes physically bound to objects or whether DNS has got a lot of people hooked on the idea that authority to mint names and authority to bind them always go together. Those two things coexist for understandable reasons at Heinz and in DNS. But the next generation of Internet naming and binding systems should separate the two.
internet & mobile systems lab hewlett-packard laboratories 1501 page mill road, ms 1u-17 palo alto ca 94304-1126 usa
firstname.lastname@example.org voice +1 650 857 5609 fax +1 650 857 2358