[rdiff-backup-users] regressing, time required

[listserv . traffic]

I have questions about regressing repo's

Environment:
Windows generally: using the win binary.
Source and dests are local disks, NTFS.

(It might be helpful to explain in generality, and then specific to
the above environment as it is different. I'm likely to add a topic
on it to the Wiki - so this is for my knowledge and also to prepare
for others...)

---
Occasionally I have a backup fail and it has to regress the
destination.

That seems to take a long time. While I don't have any specifics at
hand, I think it often take a lot *longer* to regress the repo than it
would take to do the backup itself.

Can someone explain why? (What exactly does the system do to regress
a repo?)

Can someone give me some rough estimate on times to expect?
(i.e. 2x backup length, x minutes per 10K files etc.)

Is there a way to speed this up?

---
I worry about a big repo hanging during a backup, and a regress
taking 10+ hours and running us into the middle of a working day in
some cases which could be a problem.

Thoughts, or pointers to places I can look. I don't see much detailed
info in the wiki or in the past year in the list archives (though I
could have missed something...)

Thanks,
Greg