Sister and brother dirvirts:

I began using dirvish 15 years ago.  Dirvish replaced a
vast pile of 8 mm cartridge tapes (still awaiting copying
and disposal), which in turn replaced decades of media
and format and software changes.  Dirvish produces rsync
images with enough metadata to turn those images directly
into replacement system drives;  I can still turn 15 year
old dirvish backup images (redhat 9?) into boot drives,
while the old tapes will require complex handling and
translation to recover and organize old files.  Some
are in formats that have disappeared from the world.

Dirvish author jwschultz was very helpful, and generous
with his time.  In gratitude, I wrote up his help as a
three part article for Sysadmin magazine.  When the
magazines appeared, jwschultz was already gone; we learned
of his mysterious death months later.  His parents in
Sacramento, unaware of his accomplishments and legacy,
had already wiped his hard drives ("encrypted" with a
non-windows OS) and donated his computers to charity.

I found myself (a chip designer, not a programmer) the
maintainer of dirvish, and the many organizations relying
on it (including Dutch public television and kernel.org).
I resisted major change; backup systems MUST be reliable,
elegance is optional.  Many proposed changes would have
made backups different; perhaps easier to set up, but hard
to recover in the future, or less secure across networks.

----

A month ago, I asked for help carrying dirvish into the
future, and volunteers are working on this transfer now.

I'm confident that future progress will improve usability
while maintaining compatability, perhaps with a rewrite
from "write-only" Perl 5 to Ruby or another language.

Architecturally, dirvish should remain simple and
"pull-driven", and the old config scripts should work
forever, even as new formats emerge.  Backups aren't
trustworthy if recovery requires complex translation.

Backup isn't a magic ritual; it is preparation for
unpredictable failures.  Disaster data recovery is always
hurried and inconvenient; I hope developers will focus on
streamlining and simplifying recovery to be quick, easy,
and profusely informed in very stressful circumstances.

The new developers will make those decisions.  I hope
rule number one will be stability and trustworthiness - 
followed by consensus, and respect for the needs and
abilities of a broad range of users. 

We do this to help each other.  Be good!

Keith

-- 
Keith Lofstrom          kei...@keithl.com
_______________________________________________
Dirvish mailing list
Dirvish@dirvish.org
http://www.dirvish.org/mailman/listinfo/dirvish

Reply via email to