Re: Namespace names: a semi-serious proposal

On Fri, May 26, 2000 at 03:20:03PM -0400, Paul W. Abrahams wrote:
>Michael Mealling wrote:
>>On Fri, May 26, 2000 at 02:35:27PM -0400, Paul W. Abrahams wrote:
>>>"Arnold, Curt" wrote:
>>>>Isn't that already achievable by using a URI with a protocol of "uuid:"
>>>>for the namespace and using the UUID generation algorithms.  That could
>>> >be a solution for relative URI's that are intended to indicate a document
>>>>specific namespace.
>>>
>>>Where are the UUID generation algorithms described?  I couldn't locate a
>>>description of UUIDs using the IETF search facility.  And as I asked
>>>John Cowen about OIDs, can anyone get as many UUIDs as
>>>they wish?  Can my grandmother get a bushel of them?
>>
>>Part of hte problem with UUIDs is who's definition you go with as several
>>entities have things called UUIDs and/or GUIDs. Microsoft has one
>>definition. Guid.org has another. I think ISO has yet another. Some
>>use an algorithm that includes your MAC address which raises some
>>hairy privacy concerns.
>>
>>The nice thing about most GUIDs is that they're really just a combination
>>of some location, a timestamp, and some hash which make it so that
>>you can create an insane amount of them algorithmicly...
>
>Given the multiple definitions, is there any possibility of overlap, i.e., two
>definitions might yield the same GUID?  Can you point me at the Microsoft
>definition?

Most definitly. Some are inherently incompatible. Microsoft's are
just a 16 byte number. Others contain alpha-numeric bits. So incompatible
that even charater escaping is badly handled. I'd suggest picking one
and running with it. The Microsoft one uses the definition that
comes from OSF DCE. I looked on the opengroup site but can't find a 
reference...


>And what do you think of the idea that led to this query: using UUID's as 
>the only non-deprecated form of namespace name?

I think its a bit restrictive. I'd personally use the 'urn' scheme but
that's just my personal preference. I doubt if the XML Namespace spec needs
either level of uniqueness/persistence. Others on this list seem
to want someothing that is much easier to generate and only requires
the thing to be unique to the XML document. I.e., I know what I want
but that doesn't mean I can convince the list of that....

-MM

-- 
--------------------------------------------------------------------------------
Michael Mealling	|      Vote Libertarian!       | www.rwhois.net/michael
Sr. Research Engineer   |   www.ga.lp.org/gwinnett     | ICQ#:         14198821
Network Solutions	|          www.lp.org          |  michaelm@netsol.com

Received on Friday, 26 May 2000 15:49:27 UTC