On 6 Aug, Bill Davidsen wrote:
> "Dave Platt" <[EMAIL PROTECTED]> wrote:
>
>> Here's a progress report / reaction, based on my experiments on two
>> Linux systems running Debian 2.2.
>>
>> Executive summary: Pretty Darned Cool. Thanks!
>
> If someone were aiming at a backup in minumum time vs. minimum space,
> they might lear from a product calle "FastBack" intended to backup to
> floppies. It had three modes, min time, min space, and "keep up" (by
> some catchy name). The last is of some interest, since the compression
> and CPU needed was adjusted during the backup.
>
> It turns out the gzip handles this nicely. You can gzip a file with one
> setting, gzip another and concatenate it on the end of the first, etc,
> then zcat the whole thing. I have some huge logs which I compress and
> append every eight hours, although I use the same setting for all.
>
> So it would be possible to read in blocks and compress more or less
> based on what's in a program internal buffer to send to the CD.
>
> This is a musing only, you wouldn't do it in shell unless it was the
> only thing you had, and it would be nasty to implement, so it's just
> food for thought.
>
> Thank you for the review.
>
Two comment (thoughts)
First, I have to add that the usage of the new compression mode of dump
is essential when writing to a RAW cd device. The reason is that
when writing in raw form (i.e. without an iso filesystem) the device
cannot write variable blocks (only 2Kb blocks). Therefore restore
cannot detect the correct end-of-volume.
But when using the compressed mode, each block is prefixed by a
4 byte header containing a length byte which cannot be zero.
Since cdrecord pads with zeroes this is a means to detect the
last block.
Second, by using different compression modes one can influence
the load of the processor. And writing with speed 1 but compression
ratio 3 is faster than writing at speed 2 with no compression.
--
Helmut Jarausch
Lehrstuhl fuer Numerische Mathematik
Institute of Technology, RWTH Aachen
D 52056 Aachen, Germany
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
