rsync - Feb 2006 - Fwd: Re: How to use multiple link-dest directories?

I forward Maynard's suggestions to the list.

-------- Forwarded Message --------
From: Maynard Handley <name99@name99.org>
To: Matt McCutchen <hashproduct@verizon.net>
Subject: Re: How to use multiple link-dest directories?
Date: Thu, 23 Feb 2006 19:31:56 -0800

Great. That's exactly the sort of answer I wanted. I'll try it tonight.
(BTW, I guess maybe it got lost since I was trying to be very  
explicit in my example, but yeah, I know that I shouldn't use the  
dest dir as a link src dir --- I would never even have thought that  
made sense, but I guess maybe someone somewhere has misunderstood  
things and tried that.)

The only thing left is, as I said, to give an example in the man page  
--- I trust you'll agree that it's not obvious that this is the  
correct syntax.
(And while I'm in the mood, it would be nice for the man page to  
clarify exactly what "fuzzy match" matches, and what it costs. My  
guess is that the important case it matches is a file that is  
renamed, that it operates through the construction of a hash table  
based on file-size/mod-date per directory, that the table is created  
on entry to a directory and destroyed when we move on to the next  
directory, and that assuming you have the RAM and CPU (ie you are  
throttled by your disks and/or network) it costs pretty much nothing,  
certainly no extra stat's or anything like that. But it would be nice  
to have documentation that made it clear if these are all the case.)

Thanks,
Maynard
-- 
Matt McCutchen
hashproduct@verizon.net
http://hashproduct.metaesthetics.net/

> (And while I'm in the mood, it would be nice for the man page to
> clarify exactly what "fuzzy match" matches, and what it costs. My
> guess is that the important case it matches is a file that is
> renamed, that it operates through the construction of a hash table
> based on file-size/mod-date per directory, that the table is created
> on entry to a directory and destroyed when we move on to the next
> directory, and that assuming you have the RAM and CPU (ie you are
> throttled by your disks and/or network) it costs pretty much nothing,
> certainly no extra stat's or anything like that. But it would be nice
> to have documentation that made it clear if these are all the case.)
>
> Thanks,
> Maynard

Hi there,

well my needs for fuzzy-match are a bit differing. My users are
moving files from one dir to another, mainly without renaming.
These files are about 1-5MB with a slow line, mostly a night is
to short to copy the the files again to the destination.

Can I use something like recursive-fuzzy-match or how about compare-dest?
I do not know how compare-dest exactly works, does it compare recursively?

Kann compare-dest than be the same as the rsync-destination?
Ment to scan teh whole destination tree for mathicn each source file?
Of course delete-after will be user.

Please, if anyone has any ideas, I'll try it!

Thanks in advance,
Daniel