Subject: | |
From: | |
Reply To: | |
Date: | Thu, 2 Oct 1997 12:02:07 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
James writes:
> Any conjecture/knowledge on why HP put the Jumbo's in the POSIX
> namespace instead of MPE/iX?
Yes...from the start, we said that the best place to implement
datasets > 4 GB was inside the MPE file system ... and we also
agreed from the start that such an enhancement to the file system
was unlikely to take place in the near future. So, as a somewhat
reluctant second choice, the "use more files" approach was taken.
Once you've decided to use multiple files to get beyond 4 GB,
there's really only one question: naming convention.
If we'd chosen to stay within the non-HFS (aka "ordinary MPE"
naming convention), we'd run into problems with large databases.
Already, <root>## are all used, as well as potentially
<root>A#, B#, C#, ..., J# (for up to 199 datasets), *and* some
third-party indexing products use #c (where c is a letter) ...
all in all, very few predictably unused suffixes were left!
But...MPE/iX 5.0 (yes, I'm ignoring 4.5) opened up a
great opportunity to add the extra files without *too* many
problems.
Note: the forthcoming B-Tree feature also uses HFS file names:
E.g., for a set called SALES23.PUB.DATA, the B-Tree file
will be /DATA/PUB/SALES23.idx
> Seems to me that when they
> implemented in 5.0 it [assuming "it" means: HFS file]
> was supposed to be optional. There was
> a discussion somewhere (on the list?) during the 4.5 era regarding
> putting in a "switch" to turn it off. That got shutdown for technical
The availability of HFS was never planned to be switchable on/off.
> That got shutdown for technical
> reasons which I do not recall.
--
Stan Sieler [log in to unmask]
http://www.allegro.com/sieler.html
|
|
|