Also Sprach Brian Cuttler:

> 
> Couldn't you still recover from amanda tapes if the data where
> written in 'chunksize' bits across more than a single tape ?
> 
> I mean, keep the chunks in order so you can find them, append them
> together when writing the tape to limit the amount of reassembly
> work. How are chunks re-assembled by taper now ?
> 
> Couldn't we recover with "# dd" and "# cat >>" or would reassemble
> need to be more complex than that ?
> 

Unless the tape size estimate was spot on all the time, you'd
still end up with the last chunk on a tape being a fragment
rather than a whole chunk and retrying that chunk again on the next tape.

But that's still not bad, you could check the header on the last chunk on
tape 1 and the first chunk of tape 2 and if they are the same
go with the latter. So taper would have to be rewritten to add a
header per chunk.  This would make recovery with standard tools
a bit more painful but you could still do it with dd and cat.  And
you could get previous behavior by making the chunk size the same
as the tape size.

----
C. Chan <c-chan AT uchicago DOT edu >
GPG Public Key registered at pgp.mit.edu