> If not, it is possible that my employer would be willing to pay
 > someone to bring it sufficiently up-to-date that it actually works.
 > (Since the backup image is created OK, we are expecting that this
 > ought not be a huge amount of work for someone with a reasonable
 > ability & willingness to develop in/for the win32 environment.)  It
 > would be Nice(tm) if, aas an additional result, it would be possible
 > to do an amrecover from the UNIX side against a backup created in the
 > win32 side and have a reasonable expectation that some arbitrary set
 > of the files could be restored.

It's actually rather much a pain, although this is someone who's
primarily a unix geek talking who only has marginal experience in the MS
world. (and most of it under win 3.1).  This is assuming you don't want
to break the unix-philosophy driving much of amanda's design.

I've actually gotten everything to build in client-src natively under
Windows (using visual studio .NET and a few hacked up Makefiles, no
configure), but not actually working.

I think all that's really left is getting the event driver working
and figuring out how to interact with some backup program, although
each time I get someting working, I've found something else that needs
attention.

Unfortunately, things like select() don't work the same way under
Windows as they do under Unix (you can't select on a file descriptor,
only sockets, for example).

For the software required to backup, it would be nice if amanda stored
things in the microsoft tape format used by the "backup" program that
ships with the NT Track, that is:

        http://www.layton-graphics.com/mtf/MTF_100a.PDF

I don't think "backup" can dump to stdout.  However, in theory this
would allow for bare-metal restores if you have the dump image.

I haven't really considered amrecover yet, figuring that in a pinch
someone could use microsoft's backup utility over samba to a big disk on
the unix side if something was generating dump imaages in MTF.

Also, for people just wanting to get at files in the backups without
having to touch NT:

        http://www.layton-graphics.com/mtf/

would mitigate that (I think; I've never tried it).  Certainly WinTar
(or whatever) would be another reasoable alternative.

all of my work was against the 2.5.0 track. (top of the cvs repository)
which has significantly different innards for the clients, but should
theoretically be compatible with 2.4 servers. (never tried).

I was avoiding all the glue packages lke cygwin or mingw for at least
the client portion to try to simplify the world for would-be users.

I've touched this maybe a few hours recently (it still builds) but not
with any significance since last March or so.  One of these days, I'll
get back to it.

-Todd