Bob La Quey wrote:
Would you care to elaborate on the advantages/disadvantages
of FUSE. I don't really get it.
FUSE is Filesystem in USErspace:
http://fuse.sourceforge.net/
With it you can take any sort of data you want, an RSS feed from Digg
for example, and wrap it in filesystem semantics and present it to the
kernel like a real filesystem which can be mounted. Now you can
view/manipulate your data using normal filesystem tools.
The most practical use of FUSE which I have seen so far is sshfs:
http://fuse.sourceforge.net/sshfs.html
Maybe you need access to a remote filesystem on your local box but you
can't run samba or NFS or whatever across the public Internet. You can
set up sshfs and it will ssh into the remote box and use normal unix
commands over there to manipulate the remote filesystem on your behalf
through the sshfs/fuse system which makes the remote filesystem look
like it is actually mounted to your own. I fear that's not a very good
explanation. Read the sshfs webpage and perhaps it will be more clear.
The advantage of FUSE is that you can wrap any sort of data (rss feed
from Dig, ssh pipe from a remote machine, etc) which is all stuff that
normally runs in userspace with a filesystem like interface and present
it to the Linux kernel without actually having to write a new filesystem
for the kernel. Another advantage is that if something goes wrong in the
userspace filesystem it won't bring down the whole machine.
The disadvantage is that it will be slower than running everything in
kernelspace. And there may be certain things which might not really give
you perfect traditional filesystem semantics. For example if you do a mv
of a file in an sshfs I don't know (I doubt it) if the move can really
be considered atomic. You also have the issue of userspace/kernel
context switching to consider which can slow things down. There are
surely others.
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list