[Info-vax] Backup TK50 tapes

[glen herrmannsfeldt]

Johnny Billquist <bqt at softjar.se> wrote:

(snip on tape block sizd)
>> Hmm, just like IBM. The S/360 instruction set has instructions for
>> doing signed halfword (16 bit) arithmetic, but not for unsigned.
>> (Actually, it is load halfword that is important.) One more instruction
>> would allow zeroing the high bits, but was not worth doing in the OS/360
>> days.

>> The channel hardware allows 16 bit unsigned.
 
> Well, the VAX as such have instructions for dealing with 16-bit values 
> as both signed and unsigned.
> But even if it hadn't have had that, it's just a bit of convenience. 
> You can solve that by just a few more instructions.

But those instructions were pretty expensive. For OS/360, 256K of
core was a large machine and 64K (total) a small one. The smaller
S/360 ran TOS/360 or DOS/360. At 32K, the gap overhead is reasonable.

For S/360, magnetic core memory was in the $/byte range, so those
extra instructions to allow for more larger blocks would have
been pretty expensive. (And the dollar was worth more back then.)

> So, if there really is a limit to 32K in VMS, it's in the software. 
> From a hardware point of view, the maximum block size is 64K-1.

As well as I remember, 256K was common for the early VAX, so again
why would anyone want to use more than 32K for a tape buffer?
(64K if you double buffer.)  Neither IBM nor DEC planned well for
memory prices going down as fast as they did.

I believe that z/OS can use blocksize larger than (or equal to) 32K, 
but there are still some parts of the system that don't allow it.

The 2314 disk drive used with OS/360 allows up to 7294 bytes/track,
if written as one block.  The 3330 introduced with S/360 allows 13030 
bytes/track as one block.  (Note that IBM doesn't believe in powers 
of two.)

-- glen