[Zope-CMF] Re: cmfuid

[Tres Seaver]

Jens Vagelpohl wrote:
>> I was wondering if people had opinions on making UIDs compliant with the
>> IETF draft at
>> http://www.ietf.org/internet-drafts/draft-mealling-uuid-urn-03.txt
>> instead of using a database-generated serial number.
>>
>> Benefits:
>>
>>  - No database writes necessary to generate a unique id (as long as
>>    you have probability/entropy on your side).
>>
>>  - Well-understood format due to spec.
>>
>> Downsides:
>>
>>  - Need to be careful to have a viable source of entropy on the system
>>    you're generating UUIDs on if you want to use "Version 4" UUIDs
>>    (128-bit completely random numbers).
> 
> 
> I like the standards-compliance, but so far haven't found a need for 
> UIDs or UUIDs in general.
> 
> IIRC the current implementation was meant more as a "first cut" - 
> Gregoire (who seems the main culprit), please correct me if I am wrong. 
> There was some exchange on the lists about CMFUid vs the concept of 
> UUIDs and I believe the outcome was that CMFUid does not claim to 
> generate UUIDs at all, just UIDs. I'm not sure if true UUIDs are even in 
> the scope of CMFUid.

UUIDs have disadvantages:  they can't be used to synthesize *anything* 
useful to humans (how many of you have ever had to grub through the 
Windows registry, trying to find the component which is hosing your 
system?).

The CMFUID tool defines an interface, but leaves the UID as "opaque" 
tokens.  Thus, it allows folks who (believe that they) need UUIDs to 
supply a UID generation tool which generates them, without requiring 
that folks who *don't* need them pay the (conceptual or computational) cost.

BTW, based on feedback in Vienna, I expect the Plone developers will 
release a UUID-generating replacement tool in the not-too-distant future.

Tres.
-- 
===============================================================
Tres Seaver                                tseaver at zope.com
Zope Corporation      "Zope Dealers"       http://www.zope.com