[ On Tuesday, November 16, 2004 at 08:53:51 (-0500), David Brodbeck wrote: ]
> Subject: RE: [exim] [OT] Emergency!!! Is anyone else getting this virus/wo rm?
>
> The only problem with that scheme is if even one byte of the archive gets
> corrupted, you lose all the files after that point. With individual file
> compression, you can still recover the other files.
That's actually not true -- at least not with decent compression
algorithms and assuming use of implementations which can recover the
compression stream after a damaged section, and assuming your archiving
tool is also implemented smart enough to skip damaged headers (if, as in
the worst case, that's where the damage occurs) and recover on the next
good file, and to detect integrity problems in extracted files.
> For emailing archives it doesn't matter, but this is a big issue for, say,
> tape backups. This is why writing compressed tar archives to tape is a bad
> idea -- if you have one tape dropout, or lose one tape in the set, the whole
> thing becomes useless.
You're repeating an urban myth that _really_ needs to be busted.
There's nothing new here -- these reliability and integrity features
have been basic requirements of tape backup systems even before
compression was used in such applications. Really early tape drives
with hardware compression (e.g. some very early Exabytes) couldn't
recover anything after a damaged section, but such worries are long a
thing of the past (decades!!!).
Tar and cpio, etc. can all be implemented with automatic and relatively
safe recovery after damaged headers.
The traditional backup archive formats such as tar and cpio don't use
really secure crytpo hashes to detect integrity problems (32-bit CRCs
over 5K blocks is about the best to be expected), but that's a somewhat
separate issue that needs to be dealt with regardless of whether the
archive was ever compressed or not and regardless of whether the file is
compressed inside the archive or whether the archive is compressed as a
whole.
--
Greg A. Woods
+1 416 218-0098 VE3TCP RoboHack <woods@???>
Planix, Inc. <woods@???> Secrets of the Weird <woods@???>
