From: Hubert Figuiere (hfiguiere_at_teaser.fr)
Date: Thu Mar 04 2004 - 14:38:35 EST
On Thu, 2004-03-04 at 15:50, Tomas Frydrych wrote:
> Just some more comments on the uuids:
>
> There has been a preference option NoMACinUUID for those who do not
> wish to have their MAC address embeded in the documents; I just did
> not get to producing a GUI for setting. I am not sure how strong
> people feel about this as a privacy issue: if the feelings are strong
> we could make no-MAC the default, or, if the feelings are very
> strong, we could remove it altogether (I am not greatly in favour of
> the last one).
Please, don't use the MAC address to perform UUID.
Why not perform a MD5 of the whole document at its initial save and use
that ? MD5 provide a unique 128bits checksum whois dispersion is quite
wide. This that if you get 2 documents with the same MD5 sum, then it is
likely that you'd better play and win Loto game than write Free
Software. If you really want to be sure, catenate to the file content
the current time. That'll reduce the chances even more as we introduce a
time space. I'm pretty sure that a hardware failure is more likely to
happen than a collide on these MD5 checksums.
MD5 reference:
http://rfc1321.x42.com/
> Joaquine is right; from Win2k MS no longer uses MAC generated uuid's;
> they now fill the GUID bits randomly and call this 'version 4' uuid --
> there is only one problem with this, namely, it is not an uuid, it is
> a random number.
So who cares ? You want a unique identifier, then you get it.
Hub
This archive was generated by hypermail 2.1.4 : Thu Mar 04 2004 - 14:39:29 EST