[Cialug] July Meeting: ComSec Strategies

[Morris Dovey]

On 7/20/13 11:28 AM, Don Cady wrote:
> On Jul 19, 2013 8:49 PM, "Morris Dovey" <mrdovey at iedu.com> wrote:
>>
>> On 7/19/13 8:15 PM, Tom Pohl wrote:
>>>
>>> To me, this part of your résumé is pretty darn impressive!
>>>
>>> "Designed and implemented an IPL-able routine to provide for extremely
> rapid dump and restore to/from cards on the IBM 1130. A full system and
> user library reload could be effected in less than five minutes using the
> 1442 card read-punch – a considerable reduction from the hour and a half
> previously required by the standard IBM utility"
>>
>> Necessity is the mother of invention. During debug, the stuff I was
> working on trashed the disk with every test.
>>
>> What the guys I was working with liked was that the card data was
> compressed and that the deck could be loaded at full speed even when
> backward or shuffled, with cards oriented any which way (upside down,
> turned end-for-end, etc.)
>>
> Wow, this is just as impressive as the build!
It really shouldn’t be - the last eight columns of cards (typically) 
contained a letter, or several, plus a sequence number. I used that 
information to map the card data to the disk destination - and the code 
to flip the card image horizontally/vertically was trivial. All I needed 
to do was recognize the sequence field and adjust card orientation when 
needed.

Most of the speed came from doing card and disk I/O concurrently, using 
multiple buffers (like very primitive spooling) for both devices, and 
checking to see if the disk sector into which a card’s data was to be 
inserted was already present in one of the disk buffers to avoid 
unnecessary reads.

Considered piece-by-piece, the code wasn’t impressive at all - but the 
collection of features was pure razzle-dazzle. :-D