Re: spfile vs pfile in 9i

Kenneth Koenraadt wrote:

> On Sun, 24 Oct 2004 00:21:03 +1000, "Howard J. Rogers"
> <hjr_at_dizwell.com> wrote:
>
> Point it : using a spfile does some basic validation. It does not
> validate *everything*, and I did not claim that.

You *did* claim that the spfile's "validation" would prevent the inclusion of "accident(ally) entered non-displayable signs *OR INVALID VALUES*". Which it clearly doesn't. So the real point here is that your original claim was factually wrong.

The second real point is that the use of the spfile involves no more "validation" than using the init.ora does. You'll find out your spelling mistakes or illegal values either way.

Oh, you say, the real emphasis of my first post was that the spelling mistakes get spotted "instantly", that -as you put it- "Oracle will complain immediately if anything [is] wrong". But if I am using an init.ora, I find out the same thing in much the same timescale: all it requires is that I type the command 'startup' and I'll soon enough find out whether my typing and/or spelling is up to scratch.

These are, in short, quite feeble grounds on which to build a compelling case for the spfile, quite apart from the factual tenuousness underlying it.

> Some validation is still better than no validation. Things don't have
> to be perfect to be valuable. Not for me, at least.

That your spelling of an alter system commands gets parsed as you issue it amounts to no greater quantity of 'validation' than having your spelling in the init.ora checked when you say 'startup'.

It is like saying that shutdown immediate is "better" than a shutdown abort, when in fact both cause Oracle to do precisely the same thing (kill transactions and roll them back), but merely at different times (one at subsequent startup, one at shutdown time itself). Functionally, they are identical in what they make Oracle do.

And functionally, there is no more checking of your entries with an spfile than with an init.ora, merely a difference about when *some* (not even all) that checking takes place -either at the time of altering the command, or at subsequent startup. That is a mile and half short of your original bald statement that the spfile had a "clear advantage over the init.ora: Instant validation".

Is it a worthwhile difference? For you?? For you, obviously, and you are welcome to think highly of the spfile as a result (though if you could avoid over-inflating your claims to the point that they pass facts, it would help newbies who might read this thread).

I personally think the fact that Data Guard demands one; that it sorts parameters alphabetically; and that it prevents users from writing their life stories in it by way of change history... all these are far clearer benefits and differentiators. And mentioning those doesn't involve me making any demonstrably-untrue statements about the internal workings of Oracle components.

> You seem to be quite frustrated that more average technicians buy the
> stuff from Oracle Marketing, thus praising the spfile to be
> cutting-edge technology. Then you are going to be frustrated a long
> time still, with things like workspaces, assm, bitmap join indexes
> etc...and more to come......

What are you on about? I love workspaces, and bitmap join indexes just to deal with two you mentioned. I have written here in the past about the wonders of workspace management, and I have written, again here, paeons of praise for bitmap join indexes. And both are also documented in the short book on 9i Release 2 features which was available a while back and will be again in about four days' time.

I also love ASSM, if you really mean ASSM and not ASM (which I also love). It is a wonderful way of eliminating freelist contention, and is thoroughly recommended if you are suffering from freelist contention. It does, however, have considerable overheads which I would not wish on people who do *not* suffer from freelist contention.

> Just substract 80% of the stuff from Marketing and that's it.

No, I'll subtract 100% of the marketing stuff, and make intelligent assessments of features based on my own research and the actual facts of the matter. It's more fun that way. And a good deal more rational.

HJR
>
> - Kenneth Koenraadt
Received on Sat Oct 23 2004 - 15:44:24 CDT