[RFC][PATCH] Inotify kernel API

[serue at us.ibm.com]

Quoting Chris Wright (chrisw at osdl.org):
> * Robert Love (rml at novell.com) wrote:
> > On Fri, 2005-08-26 at 11:48 -0400, John McCutchan wrote:
> > 
> > > There is nothing in inotify_watch that is useful to other parts of the
> > > kernel other than the inode, which will be passed to the callback.
> > 
> > Also, I don't see anything wrong with not exposing data structures and
> > proving access only via abstracted interfaces.
> 
> There's no compelling reason (that I see), to make it typeless.  No need
> to expose irrelevant internals, but should be simple enough to create
> a data structure that's useable and/or extensible by the client.

I agree - so long as there is no need to make it typeless, exposing the
data structure:
	1. makes it much easier for a new person to investigate and
		learn the interface (and == easier code review)
	2. iiuc, should make it easier to detect errors with sparse
		or other code checkers.

Just my pov...  I'm sure someone feels the opposite particularly about
my point (1)  :)

-serge