[Zope-CMF] Re: CMFUid isn't generating real uids

[robert rottermann]

Hi Gregoire,

I am happy with any form of UID we get.
However I think the necessary libraries exists to have an UUID generated 
on both Windows and Linux (do not know about other systems) .

I therefore wonder why we should not use them.
We will not loose anything, but gain the ability to use the same code 
as base to create relations and such.

Robert

Gregoire Weber wrote:
> Hi Robert,
> 
> 
> 
>>I think UUID should be default. If it is overkill in a given case, then it should be replaced.
>>But I never want to run the risk to hit duplicate UID's when I import some documents from one Data.fs into an other.
>>UID's are just to restricted to be worthwile.
>>
>>Robert
> 
> 
> The current policy of CMFUid is that uids get deleted on import
> (if you leave the "Remove the objects unique id on add (and import)" flag
> in 'portal_uidannotation' checked).
> 
> Thus your code has to register the object a second time a unique id. 
> This ensures that on imports no duplication happens.
> 
> I know it get's tricky if you want to import objects beeing related
> if they rely on unique ids. But this is out of scope of CMFUid.
> This is in scope of a hypothetical CMFRelation product managing that.
> 
> See 'portal_uidannotation' (UniqueIdAnnotationTool.py) for this.
> 
> Gregoire
> 
> 
>