----- Original Message -----
From: "Norman Dunbar" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Tuesday, July 31, 2001 8:06 AM
Subject: RE: [ql-users] CDROM driver for Q40/Q60 and Qubide
> In the old days, (I remember when I was a lad ....) slaving was a good
use
> of unused memory. When dealing with large files, like the Quanta library
> guide for instance, it is a boon. Try a scan first and see how slowly it
> goes from start to finish. Then do another and see how much faster it is !
>
Hi Norman,
Yes I agree that slaving can be useful it's just the implementation that
grabs all the remaining free memory to do it that I was criticizing. That
decision has had knock-on effects in later versions of the OS.
> Look also at DataDesign, Archive etc - they all benefit from slaving, in
> fact everything does to some extent - memeory access being far more
> efficient that disc - thats how my Oracle databases work so well :o)
(Well,
> apart from me tuning them to 'perfecvtion' that is !).
>
As above, yes, but not too much: it needs to be controlled/controllable by
the user. Windows, of course, does it and doesn't give you any control
either, but then when did M$ ever worry about what their users wanted?
<OT>
One tip that someone from Oracle gave me years ago that seems odd at first
glance was to sort tables on their foreign key rather than the primary.
Their argument was that for primary key access actions most of the table is
likely to be scanned and so the order doesn't matter, but for joins having
the table sorted by the foreign key actually speeds the process. I don't
have any evidence to support this - what do you think?
</OT>
> I don't think Sir Clive was that far wrong - it was 1983 (84 ?) after all.
>
The decision would have been different, I think, if he had been using some
form of floppy as the access speeds were/are much greater. Some
implementations of Forth that I've used on other systems did away with
in-memory caching of more than a couple of pages many years ago.
I really liked the microdrive imaging provided by Miracle (and someone else
earlier I think).
> Now, in today's machines, where we have fast floppy discs, ZIP drives, IDE
> hard drives, loads of memory - sounds like a PC doesn't it ? - we really
> should have an interrupt drived I/O system which gets on with things in
the
> back ground - as we are used to - but allows us to continue what we are
> doing at the same time.
>
It would be really good to have some sort of background IO - you get it just
about anywhere else. Except on IBM mainframes that lock your session when
restoring MIGRAT2'd datasets (even though they're not supposed to do so).
> Lights blue touch paper, dons Nomex flame proof suit & retires to a safe
> distance :o)
>
=8-D
Rather you than me - it's too hot today!
Peter
--
Peter S Tillier [EMAIL PROTECTED]
[EMAIL PROTECTED]
Opinions expressed are my own and not necessarily those of my employer
