Multi-platform distribution format (Was: Backups [was

[Johnny Billquist]

On 2015-09-22 03:09, Jerome H. Fine wrote:
>  >Johnny Billquist wrote:
>
>>> In any case, adding and correcting the extra code was quite
>>> easy.  The challenge was to also add support for a user buffer
>>> being above the 1/4 MB boundary in a PDP-11 with all 4 MB
>>> of memory when a Mapped RT-11 Monitor was used since
>>> the controller supported only 18-bit addresses.
>>
>> This would be the Qbus controller. And that is an annoying detail,
>> yes. You need a bounce buffer in the low part of memory. One of a few
>> Qbus controllers with 18-bit addressing for DMA.
>
> Well, it was an early controller and since it was for a floppy
> drive and the problem was really only with a Mapped RT-11
> Monitor when the user buffer was in extended memory above
> 1/4 MB, it really was not a major problem.   What actually
> might be more of a problem (if I ever finish the job of placing
> the code and bounce buffer into extended memory) is making
> sure that the portion of extended memory is below 1/4 MB.
> That would be the responsibility of the installation code to
> allocate the memory and check to make sure that the bounce
> buffer is below 1/4 MB.
> I guess that could also be done with the RK05 controller
> and the original RL01 / RL02 controller.  The latter versions
> of the RL01 / RL02 controller support 22-bit user buffer
> addresses.  What about tape controllers?

Yes, it was an early controller, from before the Qbus actually was 
22-bit addressed.

RSX solves this by creating a partition when the driver comes online, 
and making sure this memory partition is below 256K. And then all DMA 
goes through that memory partition, and then the drive copies the data 
to the final destination when DMA finishes (or the reverse on writes).
All done only in the case it detects it's an RXV21 and not an RX211.

>>> Another problem was that the index hole for single-sided floppies
>>> was offset about 1/2" from the index hole for double-sided
>>> floppies.  That challenge was solved by using a DPDT switch
>>> to flip the sensors that were used on the DSD 880/30 and
>>> that supported using, as double-sided, floppies with the single-
>>> sided index hole.  While a number of 8" floppies had been
>>> purchased that had the double-sided index hole, that was less
>>> than 10% of the total and after punching the extra pair of holes
>>> in single-sided floppies just a few times, it was very quickly
>>> apparent that the DPDT switch was a much better one-time
>>> solution.  What was initially a surprise was that EITHER the
>>> single-sided OR double-sided index hole could be used with
>>> the same floppy to access the sectors even though the holes
>>> were in different positions.  The timing did not seem to matter.
>>> Only the device driver software cared if the bit was set one
>>> way or the other, so flipping the sensors which were activated
>>> was an excellent one-time solution when the user (me!!)
>>> wanted to use a floppy with a single-sided index hole as a
>>> double-sided floppy.
>>>
>>> In any case, the code was enhanced, my version of DYX.SYS
>>> supported the RX03 double-density, double-sided floppy drive
>>> under a 22-bit RT-11 monitor.  So I set about the job of the
>>> LLFs for double-sided 8" floppy media.  As mentioned above,
>>> in addition to a couple of dozen 8" DEC floppies, I had about
>>> a dozen other brands.  To make a long story short at this point,
>>> the results were "interesting".  Every non-DEC branded 8" floppy
>>> could hold an LLF for double-sided, double-density.  On the other
>>> hand, I seem to remember that only about 2/3 of the DEC 8"
>>> floppies managed to complete the LLF.  The other 1/3 of the
>>> DEC 8" floppies could hold an LLF on the normal first side,
>>> but not on the second side.
>>>
>>> Obviously this story was somewhat different since it was not
>>> necessary to ask DEC maintenance to make the LLF capability
>>> with the DSD 880/30 to work - it already worked.  In addition,
>>> there was no DEC maintenance contract in the first place and
>>> there was no 50 gallon oil drum.  There was also no refusal
>>> by DEC to enhance the DY.MAC device driver to support
>>> the RX03 floppy drive since DEC was not asked.
>>>
>>> Over the decades since, I have always wondered how it was
>>> even possible for ONLY the DEC 8" floppies to be unable to
>>> take an LLF double-sided when every other brand managed
>>> to do so.  There was probably one floppy that was so severely
>>> damaged that it would not take an LLF on either side, but that
>>> was a specific exception.  Any 8" floppy which could take a
>>> double-sided, double-density LLF held the data successfully
>>> when used in practice.
>>
>> Probably qualification differences. DEC only cared if one side was
>> good. So floppies with one bad side were still acceptable for DEC,
>> since they only used one side anyway.
>> Floppies sold as double sided needed to pass testing on both sides.
>
> I would agree with your analysis if any of the other non-DEC floppy
> brands had even a couple of floppies which had not accepted an LLF
> for double-sided operation.  BUT,  NOT  EVEN  ONE??????????
> It seems a bit unlikely unless DEC specifically managed to locate a
> shipment at an excellent cost advantage and the floppies could be
> made at less expense.

Why? Other manufacturers sold floppies the thought might be used in 
double-sided drives, so they didn't ship any that wasn't good on both sides.

> Don't forget that except for less than 10% of the floppies being
> specifically for double-sided operation, all of the other non-DEC
> brand floppies were also sold ONLY for single-sided operation.
>
> So how did ONLY  DEC manage to acquire the one batch of
> floppy media that supported using a substantial portion of them
> (about 33% if I remember correctly) on only one side even
> though they were coated on both sides.  In fact, I used to wonder
> why the floppy material was coated on both sides in the first
> place.  By then, double-sided 5 1/4" floppies were the standard
> for the PC systems, so it was reasonable for all floppy media
> to be produced to support double-sided operation.
>
> ALL  LONG  AGO  NOW  IN  THE  PAST!!!!!!!!!!!!!!!!!!

Just because they were sold branded "single side" didn't mean the 
manufacturers didn't check both sides and just trashed the ones not 
passing. It all depends on knowing you customers and the potential use 
they made of your product. Noone but DEC customers would by a genuine 
RX01 floppy in the first place, and those who did used them exclusively 
on RX01 drives, so it was pretty much a given that only one side would 
be used. Since both sides were coated, check one side, it if fails, 
check the other. No point in trashing, if one side was working. They 
charged good money for those floppies. I thought we had already 
established that.

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol