Re: Unique handle generation

[Klaus Malorny <Klaus.Malorny@knipp.de>]

Patrick wrote:
> 
> On Wed, Mar 07, 2001 at 06:22:02PM -0500, George Belotsky took time to write:
> > Someone on this list talked about the possibility of people changing
> > email addresses, and their old address being claimed by someone else.
> >
> > At first glance, an email address appears unique, but the above
> > arguments do raise a legitimate concern.
> 
> Then, just add a timestamp (of handle creation time) to the email
> address and then it is (quite) unique.
> 
> If you loose your email address, you can still use your handle, since
> if someone else creates a new handle with the same email address it
> will have another timestamp.
> 
> You can also add a registrar/registry id.
> 
> Patrick.


Although I don't know the origins of this discussion (provreg meeting?), the
big question is whether humans should use these handles directly (as I assume
from the ongoing discussion) or not. In the first case, I think handles 

  - should be syntactically simple (e.g. limited possible characters, case
    independed)
  - should be similar among each object type
  - should -- on the other hand -- identify the object type
  - should not _directly_ refer to data it describes, but may contain some
    hints

Of course a compromise is possible. For example, CORE handles contain (beside
a type id) only a number, and we got some complaints about the handles not
having a mnemonic feature. Nevertheless, having a complete e-mail address
contained in the handle is too specific, too complicated (do we have to refer
to RFC822 or similar?) and too unhandy.

regards
Klaus Malorny