[Simh] VAX Tape Emulation?

Johnny Billquist bqt at softjar.se
Thu Jan 25 19:15:43 EST 2018


On 2018-01-25 22:31, Clem Cole wrote:
> 
> 
> On Thu, Jan 25, 2018 at 4:14 PM, Larry Baker <baker at usgs.gov 
> <mailto:baker at usgs.gov>> wrote:
> 
>     Because UNIX tapes and tape driverscannot write variable-length
>     blocks and do not allow skipping forward over records between read
>     operations
> 
> 
> ​That's not a correct statement.   UNIX does it great - I've used it 
> (and rmt) for years.  UNIX's tape driver is the most flexible of any OS 
> I have ever used (and I have done tapes on probably greater than 25 
> different OS's over the last 40 years).  That's why I suggest it.  I 
> wonder what the issue for VMS is?

Agreed. Unix have no problems with skips or variable length records. But 
it does require that you use the raw device (however, no sane person 
would ever use the block tape device, that is something completely 
useless in Unix).

> Hmm.. thinking about it, TCP sockets will remove the record orientation 
> (TCP creates an stream of octets without any records), so if you are 
> doing the funky RMS like stuff, its TCP that dropping those boundaries 
> not the UNIX tape driver.   If you want same then it need to be set as 
> meta-data.  But the UNIX tape drive works as expected.   The 'block' 
> size is the size of the write, although I think the TM11 requires a 
> minimum of 512 bytes.   Records passed with the MTIO ioctl' which mt/rmt 
> all support..

This is obviously a limitation of the rmt protocol, which makes the 
assumption that all records are of equal length. Works for the standard 
Unix tools, which also tend to want to use tapes with a fixed record length.

I thought that maybe rmt did preserve record length information, but 
thinking about it I can't actually remember that it did. But in case rmt 
do handle variable length records, then it should work. Whatever VMS 
does involving RMS and so on, is irrelevant. When the data hits the tape 
there is only variable length records and nothing else to it.

And no, TM11 controllers do not have a minimum length of 512 bytes.

> That said,  I agree if he can get DECnet to work, that is likely to 
> better integrated into the utilities.

I'd dump the tapes to local files using some tape copy program. Then I 
can transfer those files around using DECnet or TCP/IP, or whatever, and 
then write them out on simulated tape on whatever other VMS system he has.

	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


More information about the Simh mailing list