Re: [rdiff-backup-users] Ignoring resource forks

[Simon Hobson]

Nick Semenkovich wrote:
> I'm running rdiff-backup on an OS X machine which seems to have an
> enormous number of resource forks, making backups very time intensive.
>
> Even when zero files change, the backup takes ~5 hours, seemingly
> because of resource forks alone. rdiff-backup produces /enormous/
> mirror_meteadata.snapshot files (~7 GB), and the diff creation in
> rorpiter.py takes forever.
>
> Is there a way to disable resource fork backups?

In general, you do not want to do that. For some files it will be OK, 
but for many more it will make the file useless.


Warning, this is from memory, and could well be wrong !

It sounds like the same problem with Apple's version of rsync (I 
guess rdiff-backup uses the same code ?) in that the way it's done 
means that resource fork* gets copied every time - it's annoying as 
it stops me doing some stuff across the internet. There's discussion 
of it on the internet, I vaguely recall having looked it up a while 
ago, but IIRC it goes something like this ...

To make the HFS+ files compatible with different filesystems, files* 
are converted to AppleDouble. This splits them into a data file 
containing the data fork, and a separate file (._<filename> ?) 
containing both the resource fork AND the metadata for the file. If 
the recipient rsync also has the Apple patch, it will reconstruct the 
file at the destination, otherwise it remains as a separate file.

I think the problem, from observation of effect (with MacOS at each 
end), not from reading any code is that the data file is processed, 
and the recipient is too dumb to realise that the other half will 
come later - so it copies the data file (which is different) and the 
other half gets deleted. Then the other half comes along and gets 
copied as well. Or it could be that the metafile comes along first, 
it isn't present on the destination and so it gets copied (resulting 
in a modified file), and then the datafork comes along and needs to 
be copied to gets things back to normal. Hope that all makes some 
sort of sense - it's a fair while since I looked into it.

* I don't think it's as simple as "files having a resource fork". 
Many filesystems don't support all the metadata that HFS+ does, and 
so Apple came up with a couple of file types to deal with it. I think 
OS X uses AppleDouble to store files on a foreign filesystem - this 
combines the resource fork with the metadata into a separate file, 
hence loads of files that start with ._

--
Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.