[PATCH 0/2] resource control file system - aka containers on top of nsproxy!

[Sam Vilain]

Paul Menage wrote:
>> In the namespace world when we say container we mean roughly at the level
>> of nsproxy and container_group.
>>     
> So you're saying that a task can only be in a single system-wide container.
>   

Nope, we didn't make the mistake of nailing down what a "container" was
too far before it is implemented.  We talked before about
containers-within-containers because, inevitably if you provide a
feature you'll end up having to deal with virtualising systems that in
turn use that feature.

> My patch provides multiple potentially-independent ways of dividing up
> the tasks on the system - if the "container" is the set of all
> divisions that the process is in, what's an appropriate term for the
> sub-units?
>   

namespace, since 2.4.x

> That assumes the viewpoint that your terminology is "correct" and
> other people's needs "fixing". :-)
>   

Absolutely.  Please respect the semantics established so far; changing
them adds nothing at the cost of much confusion.

> But as I've said I'm not particularly wedded to the term "container"
> if that really turned out to be what's blocking acceptance from people
> like Andrew or Linus. Do you have a suggestion for a better name? To
> me, "process container" seems like the ideal name, since it's an
> abstraction that "contains" processes and associates them with some
> (subsystem-provided) state.
>   

It's not even really the term, it's the semantics.

Sam.