[uf-discuss] Cross-network identification (Was: OpenID)

[John Panzer]

Scott Reynen wrote:

> On Feb 20, 2007, at 6:56 PM, John Panzer wrote:
>
>> Scott Reynen wrote:
>> ...
>
>   Here's the purpose of UID from vCard:
>
>> To specify a value that represents a globally unique
>> identifier corresponding to the individual or resource associated
>> with the vCard.
>
>
> So if two hCards have the same UID, they must refer to the same  
> person, because otherwise it wouldn't be globally unique.  And I  
> think that solves the problem Thom described above, unless I'm  
> missing why it needs to be specific to OpenID.

Nope, it solves it.  OpenID URLs are just one example.

> ...
>
>> I'm thinking here of cases where the target may be an OpenID, but  
>> not necessarily provide an hCard.
>
>
> UIDs do not need to point to hCards.  Nor do they need to be  
> OpenIDs.  They can do both, but the only requirement is that they be  
> globally unique and correspond to the subject of the hCard.  And that  
> minimal requirement seems to be just enough to solve this problem  
> without worrying about what, if anything, is on the other end of the  
> UID.
>
Thanks.  (The thread in the archives was somewhat twisty; the summary is 
helpful.)  So, rel="url uid" it is.

I think this solves the problem of how to match one hCard up with 
another using a convenient unique key well above the 80% level.

One minor point: URLs are unique but not truly persistent.  Due to URL 
reuse,when trawling through archives you can't assume that UID1 == UID2 
means person 1 == person 2 unless both UIDs were minted at the same 
time.  I'd probably solve this heuristically if ever necessary, but:  Is 
it worth an implementor's warning somewhere?

Interestingly, the vcard people seemingly dealt with this issue, because 
they edited [1]  "persistent, globally unique identifier" prior to the 
final RFC draft.

-John Panzer
[1] http://www.imc.org/pdi/vcard-21.txt