The design specification for this was discussion on pkg-discuss and at a
face-to-face meeting with Stephen and Danek in MPK. The link to the
design is in the issue:
http://mail.opensolaris.org/pipermail/pkg-discuss/2008-May/003577.html
The beginning of this thread was pointed to in comment #1 of issue
1347. I added a comment to point to the as-implemented design.
The reason why UUIDs are here is to allow analysis to be done on the
server side as to what is happening in individual images. For example,
for packages a and b, how many images downloaded one or the other or
both. When there are multiple images per system, or multiple systems
behind a firewall, the IP address in the log isn't enough to answer
these questions.
The original intent was to have one UUID per image, but based on
Stephen's input, the design was changed to one UUID per authority to
avoid potential concerns about cross-authority correlation without the
users consent through a registration.
The --reset-uuid is how you get the UUID there in the first place.
Running it again or running --unset-uuid allows the user to opt-out of
the tracking that is being done.
Thanks.
Tom
[EMAIL PROTECTED] wrote:
Tom,
Please review the changes for the following issues:
1347 optional UUID per image
Here is the webrev: http://cr.opensolaris.org/~tmueller/cr-1347/
It's not clear to me why you've taken the approach that you have. I'd
appreciate it if you would explain a bit more about why UUIDs are here,
and what problem they're intended to solve.
Why is it that we have a UUID per authority? If the UUID is a unique
indentifier for the image, it seems that we should create one during
image-create and have it be the same for all authorities.
I also don't understand why we want to reset or remove a UUID, that
seems to defeat the purpose of having a uniquely identifiable string
attached to an image.
-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
