Brian Aker wrote:
Hi!
On Aug 12, 2008, at 10:53 AM, Roy Lyseng wrote:
I guess that would be up to the user. If he wants a server-managed
timestamp, then set the "auto-update" property. If he wants a
user-managed timestamp, don't.
DATETIME is a non-managed type, while TIMESTAMP is a mostly managed type.
My point is that there is no need to have separate types for
system-managed and user-managed data. Simplifies expression management,
compatibility rules and conversion rules...
In a prior system we implemented system-generated and system-managed
columns (checksum and LSN) with the type name "special", but internally
they were just BINARY(N) with a readonly tag and a system tag. However,
we could have dropped the "special" type and then had full type
compatibility with BINARY.
We have a SERIAL type as well, though it just maps to AUTOINCREMENT. I
have thought about fixing it to just be a SERIAL and not allow for human
interaction.
I would make the same argument here...
Other people have also pointed out that autoincrement is difficult to
scale, so an IDENTITY or UUID attribute is probably more interesting in
a horizontally scaled environment (like MySQL Cluster).
Thanks,
Roy
Cheers,
-Brian
--
_______________________________________________________
Brian "Krow" Aker, brian at tangent.org
Seattle, Washington
http://krow.net/ <-- Me
http://tangent.org/ <-- Software
_______________________________________________________
You can't grep a dead tree.
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
