[SLUG] RE: compression -- incremental Vs ??

[Minh Van <mle@nospam.mail.usyd.edu.au>]

> Interesting proposal, however given that any application has to allow as
> best it can for failure -- I have had occasions under winzip 7 {yes I have
> to use M$ stuff at work} where faults arise and the compression is aborted.
> If an application did what you required and incrementally compressed 1meg
> then deleted the corresponding 1meg and a fault occurred you would have no
> original data to recover from.
> 
> Therefore, while it is I suppose it is potentially possible, I would
> consider it unlikely to be found as no programmer would want to explain to
> each and every "AOL" user on the planet why their original data was lost
> from a failed compression. 
> 
> so -- Possible but not probable and I have not heard of any such app.

a checksum of some sort could be used to match & append at points of
failure during compression. the compression program would be smart enough
to decrypt/encrypt a file "as it goes", so no matter where a file was
aborted there's enough data (checksum) to resume decompression of the file
fully & properly. 

temporary files that "buffer" information could be used during a
compression process to ensure that broken pieces could be put back
together after an abortion.

here's my illustration of a broken, aborted or failed compression.

original file:

######################

broken file:

=======---  <broken section> ---###############
(compressed)                 (to be compressed)

a temporary buffer file would store a couple of kilobytes at each end of
the aborted compression. 

"---" is the buffer ends. info in this area of either side is expendible.

==---                 ---###
(compressed end)      (to be compressed)

now all the checksum has to do is match these points, kill off any parts
of the ends that don't make sense due to the unexpected failure .. and
resume compression. so the "buffer" is in a sense a spot-on backup of the
original file. so only a section of original file is compressed at any one
time for safety, until the whole file is fully compressed.

i've got no idea how to program something like this. i'd say it'd be
pretty hard :). it's just my proposal.


--
SLUG - Sydney Linux Users Group Mailing List - http://www.slug.org.au
To unsubscribe send email to slug-request@nospam.slug.org.au with
unsubscribe in the text