Random Access Tape 2454

If you could handle error recovery on block reads right you could just keep reading. You would get one or two empty ones (the file marks marking the "new" end of tape) then one to around five erronous or "short" blocks, and it would sync right back into the file again.

Some tape drives allowed the writing of explicitly long gaps, so you could write spacers to guard against just this thing. Primos MAGSAV did such stupid tricks.

Any drive that could handle seismic tapes would handle this. The seismic tapes had no inter-record gaps. I wrote handlers for these for several OSes. The trick was to set the device up for asynchronous I-O, and start one I-O with a largish block size, and start another right away that would wait for the first one to finish with an overflow, and then restart one more I-O while the second one worked its way to an overflow etc.

DDS tapes has lots of ideosyncracies, but this is correct. You could append data by going to the end. You just had to labour a little more to do the append.

With helical tape, when you have lost sync with the track it is a lot harder to regain sync with the data. That is probably why this is a problem.

And, all these helical tapes, DDS-DAT, DLT, Exabyte and whatnot really make good ole 9" tapes look good.

-- mrr