30 May
2014
30 May
'14
2:38 p.m.
IBM famously introduced its CKD disk file format in the 60s; did anyone else
(other than IBMs clones including RCA) support such a disk format?
Specifically a disk format with a block (record in IBM terms) having a
variable length key field (including 0 length) and a variable length data
field. To the best of my knowledge everyone else used a fixed block (sector
to most) size on a disk volume but my knowledge of most of the BUNCH is
limited.
Tom
31 May
31 May
9:14 a.m.
On May 30, 2014, at 5:38 PM, Tom Gardner <t.gardner at computer.org> wrote:
IBM famously introduced its CKD disk file format in
the 60s; did anyone else
(other than IBMs clones including RCA) support such a disk format?
Specifically a disk format with a block (record in IBM terms) having a
variable length key field (including 0 length) and a variable length data
field. To the best of my knowledge everyone else used a fixed block (sector
to most) size on a disk volume but my knowledge of most of the BUNCH is
limited.
I remember that from the 360s. That?s the only place I ever saw it. Variable length
records, varying even from one record to the next, egads. You could have 80 byte sectors
storing card image files, though you were encouraged to use longer sectors for efficiency.
Lots and lots of documentation about this in the JCL manual, and utilities for dealing
with format changes, and large amounts of complexity in the OS syscalls to specify all
that. It never was clear to me what the point was.
IBM didn?t do that on other machines, though. For example, the 1620 had 200 digit
sectors, if I remember right.
Funny sector sizes were found in other machines of that era as well; the CDC mainframes
had sectors of 322 12-bit words, which were usually exposed to the programmer as 64 60-bit
words, reserving the first 24 bits for file system control purposes. (PLATO, early on,
did make all 322 words visible because it had its own file system, but at some point that
became a nuisance and there was a whole system file conversion from 322 to 320 word
blocks.)
paul
4:29 p.m.
On 05/31/2014 09:14 AM, Paul Koning wrote:
IBM didn?t do that on other machines, though. For
example, the 1620
had 200 digit sectors, if I remember right.
And a group mark, so sectors were pretty much anything from 1 digit to
200 digits in length.
Funny sector sizes were found in other machines of
that era as well;
the CDC mainframes had sectors of 322 12-bit words, which were
usually exposed to the programmer as 64 60-bit words, reserving the
first 24 bits for file system control purposes. (PLATO, early on,
did make all 322 words visible because it had its own file system,
but at some point that became a nuisance and there was a whole system
file conversion from 322 to 320 word blocks.)
Was it the 853 disk drive that was set up to take either IBM-compatible
hard sectored packs as well as the 6000 version of the packs with
different sectoring? I seem to recall a 1SP overlay that would allow
853 IBM-type packs to be used on a 6000, but with a considerable loss of
capacity.
Very dim memory of this, anyway.
--Chuck
1 Jun
1 Jun
8:48 a.m.
On May 31, 2014, at 7:29 PM, Chuck Guzis <cclist at sydex.com> wrote:
I?m not familiar with those ? only worked on 844 drives, which DEC users know as the
RP04.
The format of those packs was established by a low level micro-engine in the disk
controller (7054). I suspect that could handle a wide variety of formats just by
reprogramming the ?controlware?. Should take a look; I have a copy (but no longer any
manual for that processor).
paul
On 05/31/2014 09:14 AM, Paul Koning wrote:
Yes and no. Yes, the group mark would act as the logical end of data mark, ending the
read instruction. But the physical sector size was always 200 digits.
By contrast, the 360 disk format would actually write physical blocks the size you asked
for.
IBM didn?t do that on other machines, though.
For example, the 1620
had 200 digit sectors, if I remember right.
And a group mark, so sectors were pretty much anything from 1 digit to 200 digits in
length. Funny sector
sizes were found in other machines of that era as well;
the CDC mainframes had sectors of 322 12-bit words, which were
usually exposed to the programmer as 64 60-bit words, reserving the
first 24 bits for file system control purposes. (PLATO, early on,
did make all 322 words visible because it had its own file system,
but at some point that became a nuisance and there was a whole system
file conversion from 322 to 320 word blocks.)
Was it the 853 disk drive that was set up to take either IBM-compatible hard sectored
packs as well as the 6000 version of the packs with different sectoring? I seem to
recall a 1SP overlay that would allow 853 IBM-type packs to be used on a 6000, but with a
considerable loss of capacity.
9:37 a.m.
On 06/01/2014 08:48 AM, Paul Koning wrote:
Yes and no. Yes, the group mark would act as the
logical end of data
mark, ending the read instruction. But the physical sector size was
always 200 digits.
True, but it shows that there was a precedent for recording
variable-length records.
By contrast, the 360 disk format would actually write
physical blocks
the size you asked for.
I don't see much advantage to this--it would seem to make effective
packing of sectors on a track.
I?m not familiar with those ? only worked on 844
drives, which DEC
users know as the RP04.
I remember when the 844 was released--they replaced the very troublesome
808-sized 821 drives we were using. I don't think the 821 saw any
widespread distribution outside of Special Systems. They were an utter
disaster.
We used the 844s with a MAC and an expander. One of our customers had
144 of them on site, hooked to 4 Cyper 74s, coupled together with 4MW of
ECS. Lots of QSEs in that one.
The 852 was 1311-compatibe; the 853 was sort of 2311-ish. The packs
would fit either drive and were hard-sectored. You could use them in
"track-at-a-time (unsectored)" mode or in sector mode. They were not fast.
If Spence Preston is still around, he'd know just about everything there
is to know about 844s--he spent a lot of plane time putting out fires on
those.
--Chuck
3857
days inactive
3859
days old
4 comments
3 participants
participants (3)
-
cclist＠sydex.com -
paulkoning＠comcast.net -
t.gardner＠computer.org