Re: db cache size in oracle 9ir2

"Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message news:1069545738.980498_at_yasure...
> Howard J. Rogers wrote:
>
> > "Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message
> > news:1069539193.545809_at_yasure...
> >
> >>Come on Howard there are two other things the SPFILE is good for.
> >>
> >>1. On-line permanent parameter changes
> >
> >
> > Achievable by remembering to edit your init.ora. Quite why 'alter
system'
> > plus edit init.ora is any worse than 'alter system and remember to
> > scope=spfile if you *don't* want it to be permanently changed', I can't
work
> > out.
>
> I'll give you this one. But only because of the intrinsic logic of your
> arument.

I like to try ;-)

>
> > In other words, all the spfile does is shift the 'burden of proof', as
it
> > were. In the old days, I could dynamically alter something and be
assured
> > that such a change would NOT automatically become permanent. Now, I have
to
> > explicitly ask it not to be.
>
> And that isn't an improvement? ;-)

Nope.

Call me old fashioned. But 'alter system' has always meant 'change something about the way my instance behaves', just as 'alter database' usually means 'change a setting in my control file'.

Suddenly in 9i, 'alter system' takes on a whole new meaning: don't actually change anything about my instance, but just do some binary file editing. An alter system command that doesn't actually alter the system? Not good.

It's a bizzarre piece of syntax, and (IMHO) poor. Since they gave us every other imaginable xxxCTL in 9i, why they didn't give us an SPFILECTL utility, I'll never know. And if they were going to alter the behaviour of 'alter system' syntax, there should have been an init.ora parameter called 'INIT_SCOPE=' which was capable of accepting values 'both', 'memory' and 'spfile' so you could choose whether to sign up to the new lunacy... er, I mean orthodoxy. And make it session-dynamic so different users with different preferences can peacefully co-exist on the same database.

>
> > But when they've done all of that, you have to ask why you need an
spfile at
> > all. If you are going to store parameters in a binary file, why not
stick
> > them in the control file itself and have done with it? Guess what I
reckon
> > will happen around 13.5z??
> >
> > Regards
> > HJR
>
> Personally I'd like to see the control files go too. I still haven't
> figured out why we need them as separate files.

Because it doesn't store any real user data, and therefore has a completely different function from the data files and the redo logs?

Regards
HJR Received on Sat Nov 22 2003 - 18:22:53 CST