[Info-vax] Backup TK50 tapes

[Johnny Billquist]

On 2013-02-24 19:06, Bill Gunshannon wrote:
> In article <qpydnQZXTdEypLfMnZ2dnUVZ_uCdnZ2d at giganews.com>,
> 	drb at ihatespam.msu.edu (Dennis Boone) writes:
>>> While i'm sure there are more modern utilities to copy tapes, providing
>>> you know the block structure of the tape, dd should copy it verbatim,
>>> even if you have to write a script to read it block by block.
>>
>> NO NO NO!
>>
>> I love dd and all, but it THROWS AWAY BLOCK STRUCTURE and leaves you with
>> a byte stream.  The block structure is important on tapes.
>>
>> Don't use dd to image tapes!
>
> Unless you do each file individually.  :-)

You are lucky if you even get away with that. There is no guarantee that 
all blocks even within one file are of the same size. In fact, if it is 
an ANSI format tape, you are guaranteed it is not.

Why can't we just admit it that in this specific instance, Unix just 
don't work that well. Unix concept of I/O just being a stream of bytes 
is a nice concept in some ways and at some times. However, it sometimes 
breaks in the face of reality. I'm sure noone here have ever even used 
the block device for tapes in Unix. It is totally useless, but it is the 
layer that implements the stream of bytes even for tapes. It's just that 
people realized they don't actually want that in this case.

So everyone uses the raw device for tapes, to the point that they don't 
even realize that there is both a /dev/mt0 and /dev/rmt0.
(I wonder if Linux even have the block device for tapes?)

> That was how the Ultrix-11 boottape got trashed and stayed that
> way until correct bl;ocking could be rediscovered and reimplemented.

Unix is just so broken in some ways, it will never be fixed.

	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