[PATCH 02/14] spufs: fix local store page refcounting

[Al Viro]

On Wed, Dec 07, 2005 at 09:19:28AM +1100, Paul Mackerras wrote:
> Think about someone changing the VFS layer interface and fixing up all
> the filesystems to accommodate that change.  That person is doing some
> of your work for you, so you want to make it easy for him/her to find
> your filesystem.  That's the sort of thing I was referring to as
> maintenance.

FWIW, I think it's not a serious argument.  Interface changes => grep time.
And that means grep over the tree anyway.
 
> As for changes on the cell-specific side, the people doing those
> changes will know where to find it, so it isn't a problem having it in
> fs/.
> 
> Having it in fs/ also means that it is more likely that people
> familiar with VFS internals will look through your code and comment on
> it.  I know that can be painful in the short term, but in the long
> term it will lead to better code.

That's solved by asking for review...

As far as I'm concerned, the only thing here that looks like a possible
reason to move the entire thing is highly unusual semantics of final
close and interesting use of VFS interfaces in spu_create().  I.e. it's
not that we have a filesystem there.

OTOH, if you go looking for analogs as far as unusual interaction with VFS
is concerned...  net/unix is unlikely to get moved.