31 Aug
2005
31 Aug
'05
11:33 a.m.
>>>> "CRC" == CRC <technobug
at comcast.net> writes:
CRC> On Wed, 31 Aug 2005 09:53:55 -0400, Paul Koning
CRC> <pkoning at equallogic.com> wrote:
> I don't know about Modcomp, but that's not
the IBM way. IBM had
> amazingly primitive file systems (at least through OS/360), but
> not quite *that* primitive. You could create files at anytime,
> but only through JCL. (Actually, I think it was possible to
> create files under program control, but not via the standard
> APIs.)
>
> [...] paul
>
CRC> You are talking about big iron here. The IBM 1800, SEL, and
CRC> Modcomp were essentially toys which used a preallocated file
CRC> system to ensure speed in file access. All files were guaranteed
CRC> contiguous and of know size. The editor, compiler, and linker
CRC> each had their own segments which they individually managed. The
CRC> editor would create "files" within its segment. The JCL on these
CRC> machines was quite rudimentary and did not include any
CRC> capability to re-hack the file system. You could always design
CRC> an application to partition and manage one of the preallocated
CRC> segments.
CRC> I remember having to re-sysgen the system numerous time while we
CRC> were designing the application in order to optimize the file
CRC> sizes.
Interesting.
DEC did it much better in RT-11 -- a very fast and simple file system
without any gratuitous pain in the neck operating limitations like
that.
paul