On Fri, Nov 23, 2012 at 06:32:15PM +0100, Thomas Vander Stichele wrote:
> Heh :)
> 
> well, my goal was more to provide you guys with the tools to get the
> github repo set up as you want it, before patches come in.
> 
> I don't know who currently maintains dirvish, or who should be hosting
> the actual repo, but it shouldn't be me.
> 
> Any takers ? Any comments ? Anything you want to change in the repo
> conversion ?

Hello there, I host the dirvish.org website,  which has a subversion
server.   Dirvish is pretty much how jwshultz left it,  with a few
excellent community patches rom a few years ago.  Not many patches
are coming in these days, and sadly, most are intended to change
(aka break for others) the behavior, not fix defects.

Whither dirvish?  Well, some folks want to add secret sauce and
sell it as proprietary - sorry, jwshultz provided divish via the
open source license, it's not BSD so you can't extend it that way.
It would not be difficult to completely replace with new code.
rsync remains GPL, and the foibles of Windoze and Mac can limit
the usefulness of rsync-based software on those platforms.

Some folks want to change the functionality, but without testing
the impact on large existing operations.  For example, dirvish
is used for backups (with major mods) by the Open Source Labs
at Oregon State University, which hosts the linux kernel,
mozilla/firefox, and drupal, along with many other important
open source projects.  I think it would be great if dirvish
became even more useful to them, and their patches were available
to us, but we must not make their backup or restore process fail.

Sometimes, folks want to make big changes like transforming dirvish
into a "push" backup system.  That particular change might be more
convenient for some, but mostly convenient for system crackers 
looking for an easy way to hack into the backup server and then
subsequently all the machines for which it contains security
information in the backups.  Sorry, but pull backups make far
more sense security-wise, though we can develop handshake methods
that make the initiation, interruption, and continuation of 
backups more intelligently scheduled.

All that said, if anyone wants to create new versions of dirvish
and support them on github or other server, I would be glad to
link to them.  I will even provide space on the main website to
describe intended use.  I will insist that both the advantages
and drawbacks are included in the descriptions, including pointers
to longer descriptions (perhaps on the wiki) of how the changes
were tested.

I would especially like to see improvements in installation,
testing, validation of existing backups, and most especially
restores.  Restores are the weakest link - while automated
backups are dandy, system restores are the main reason to make
backups in the first place, and system restores are never easy,
usually needed during unscheduled disasters.  I wrote my own
adhoc scripts for this, and can generally restore one or two
machines in an afternoon, but my hacks are not generalizable;
and I don't do fire-drill testing of them nearly often enough.

There is indeed much room for improvement, and a git repository
would be a good way to get that started.  But please keep in
mind the installed base, and the consequences of "improvements".

FYI, I migrated the wiki to Moin, and can give access to others.
The wiki has been abused in the past, so I am judicious but not
miserly in my trust.

Keith
-- 
Keith Lofstrom          [email protected]         Voice (503)-520-1993
_______________________________________________
Dirvish mailing list
[email protected]
http://www.dirvish.org/mailman/listinfo/dirvish

Reply via email to