[Info-vax] Backup TK50 tapes

[Johnny Billquist]

On 2013-02-27 14:07, Stephen Hoffman wrote:
> On 2013-02-27 03:10:49 +0000, David Froble said:
>
>> I'm going to take a guess here and state that the 32K limit was in
>> VMS.   Specifically, there is a VMS limit of 32K for string lengths.
>> The string descriptors use 16 bits to specify the string size.  I'm
>> going to guess that string descriptors might have been used to specify
>> buffers. 16 bit length, 32 bit address ...
>
> There was a limit of 32 KiB for string descriptors, and that old limit
> is still present on OpenVMS VAX systems.  The limit was lifted on the
> 64-bit systems at OpenVMS Alpha V7.0, and on OpenVMS I64 systems. Quick
> C example of 64-bit string descriptors:
> <http://labs.hoffmanlabs.com/node/787>

One reason for the restriction on VAX might be because string 
manipulating instructions actually have a 16-bit field for string 
lengths, meaning the maximum string length for any VAX instruction is 
65535 bytes. Which is one less than 64K. So, a limit of 32K might not be 
totally unreasonable in software.

> (Whether it was the size of the tape controller hardware registers or
> maybe ANSI tape specs or the old 32-bit descriptors that led to the
> transfer size limit or a general "32-bits is good enough" moment,
> donno.  Don't really care all that much about the origin of the limit,
> either, as I'm not booting off tapes, not copying tapes, and have
> relegated the remaining tapes to long-term archival-storage purposes,
> for those cases where there are still tape loaders or tape libraries
> around.  I'll leave that to the retronauts and retrohistorians to debate.)

I assume you meant "16-bit" when you wrote "32-bit" above... :-)

	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