Re: Oracle 9i DB Security Hole

In article <3CC00A28.A05CFBBD_at_ca.ibm.com>, you said (and I quote):
> If Oracle's compiler logic is anywhere close to how DB2's works I can very
> well understand how such a bug occurs and reaches production.

how about worse? ;-)

> It is not
> obvious that adding a new SQL operator in a parse tree (or Query Graph Model
> in the DB2 case) can impact authorization checking for base objects.

IIRC how these things were done (back in the days when I had access to source code of db s/w...), the parse is independent of the access check. One cannot access-check something that hasn't yet been recognized as "accessible". It used to be done at p-code generation. Ie, the step where the parsed command is translated into a series of p-code commands that then are checked for validity/security/whatever. Then the whole lot is sent to the optimizer. Then and only then does it run. Of course, this was more than a decade ago. So take it with a grain of salt: wouldn't surprise me if someone just came up with an "optimization" that allows this sort of rubbish to creep.

>
> I'd be the last to throw a stone on that one. Sooner or later it may be IBM's
> or (count my blessings) even my turn.
>
> Nonetheless Oracle would be wise to retire their "Unbreakable" marketing
> campaign.

You are absolutely right. Never liked it myself. It's the typical arrogant marketing statement that gets a company in hot water and destroys reputations.

> Mortality is just too obvious now and the competition will likely and happily
> point that out in case they forget about it ever again :-)

You bet!

-- 
Cheers
Nuno Souto
nsouto_at_optushome.com.au.nospam