> On Apr 12, 2024, at 2:28 PM, Paul Berger via cctalk <cctalk@classiccmp.org>
> wrote:
>
>
> On 2024-04-12 2:45 p.m., Christian Kennedy via cctalk wrote:
>>
>> On 4/12/24 10:28, Chuck Guzis via cctalk wrote:
>>> Isn't that the IBM 2321 Data Cell drive?
>>
>> Same idea, but I recall the cabinets being lower to the floor and the media
>> being more rigid than the 2321 noodles. Then again, it's been the better
>> part of 50 years, and it could well have been a 2321.
>>
>> Memory rot sucks.
>>
>>> Having one's files "photostored" at LLL was a chancy proposition. There
>>> were bootleg programs to access every file for a user, just to keep them
>>> from being consigned to the photostore.
>>
>> It was chancy at LBL as well. The mechanical handling of the 1360
>> photostore cells was something that would have defied the imagination of
>> Rube Goldberg, and chips routinely ended up in places where they didn't
>> belong (although they did make pretty cool bookmarks for my teenage self).
>>
> The problem with a lot of these old machines was they relied on a lot of
> electro-mechanical devices that would today be replaced by electronics and a
> few simple actuators. These mechanical devices need to be adjusted and
> maintained and have lots of parts to wear out. While I only started with
> IBM in 1979 I still got to work on machines that would now be considered
> electro-mechanical nightmares.
Some of the earliest magnetic storage was mechanically simple: magnetic drums.
Nothing moving apart from the spinning media, and quite fast. Fixed head
("head per track") disk drives are a variation on that theme, DEC had some that
were successful for a while.
I remember a concept for a very fast magnetic storage system that didn't become
a product, as far as I know. The scheme was to build a large array of heads,
using IC-manufacturing type techniques, and mount that array in contact or
near-contact with a flat rectangular magnetic plate. The plate (or the heads)
could move a small amount in one direction. The idea was "head per sector",
with the mechanical motion scanning the sector across the head. Given
something like piezo-electric actuators it would have been quite fast.
There's a neat document in the CWI archives, a course on computer design from
early 1948. It has a section about memories, well before core memory was
invented. The schemes it describes are quite curious, including photographic
memories, selectrons, and various other schemes. Also drum memories, including
the rather mythical notion of a drum spinning at 60,000 rpm.
paul
