> Now, how will we work? I have some ideas about this, which I'd like to
> discuss with everybody here. I think that the UML patch will become
> the "development tree" of UML, where new patches hang around to get
> some testing, while the mainline kernel should get only safe patches.
> For instance, for now the humfs and filehandle work must *not* go in
> mainline, since they are highly experimental.
Yeah, this sounds reasonable. My tree will be experimental stuff, and
forwarded to Andrew/Linus when it is considered OK.
> Also, the various uml patch should never be actually "merged" inside
> the UML patch: it must always be a patchset, i.e. a collection of
> separate patches. Like -mm. So merging with mainline will be easier,
> as reviewing separate patches or dropping bad ones.