[Gllug] Network traffic

[tet at accucard.com]

>With that in mind, there is no a priori reason why the nfs couldn't
>be written in such a way that the remote mount sendt a 'cp' like
>instruction to the server nfs when it knew the operation would result
>in redundant network traffic.

In principle, yes, that's possible, but it assumes the filesystem/vfs
has a "copy a file" API, which isn't true. Instead, it just has simple
operations like read, write, mmap, etc.

While it would be possible to have a copy_file API that took account
of network location to try and minimize network traffic, it's not as
trivial as you might think. It couldn't be done directly at the
filesystem level, for example (think about what happens if you want
to copy between two separate filesystems on the same remote machine,
or even on a local machine, for that matter). So you'd need to have
something at a higher level implementing a copy mechanism. But at
that point, you're just bloating the kernel with things that really
belong in user space.

Tet

-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug