[cap-talk] Domain change (IPC?) overhead

[David-Sarah Hopwood]

Jonathan S. Shapiro wrote:
> On Tue, 2008-03-25 at 11:50 -0400, Jonathan S. Shapiro wrote:
>> On Tue, 2008-03-25 at 07:52 -0700, Mark Miller wrote:
>>> On Tue, Mar 25, 2008 at 6:30 AM, Jonathan S. Shapiro <shap at eros-os.com> wrote:
>>>>  I think you are tacitly assuming that domain calls will be normal
>>>>  procedure calls. No sensible design would do this.
>>> Unless I misunderstand your meaning, Gedanken, W7, E, Emily, Joe-E,
>>> Caja all do this.
>> Yes. And I stand by my assertion. The domain interdependencies resulting
>> from these systems do not scale up successfully.
> 
> Hmm. I didn't intend to preclude the possibility of domain boundaries
> that might exist at multiple granularities. A single domain boundary
> need not provide all forms of isolation.

As is the case in E, for example. A message send between E vats is not
strictly speaking a procedure call - it is very similar, but the isolation
provided between vats is greater than between objects.

>> For example, it may be
>> acceptable for one domain boundary crossing to share a space accounting
>> realm, but not for another.

If, for the sake of argument, a system were designed to provide space
accounting at vat granularity but was otherwise identical in semantics
to E, would that system "scale up successfully", in your opinion?

-- 
David-Sarah Hopwood