[cap-talk] Capability accounting

[Jed at Webstart]

At 07:15 PM 6/22/2006, Norman Hardy wrote:

>On Jun 22, 2006, at 6:48 PM, Nick Szabo wrote:
>
> >
> > The main barrier to micropayments and agoric allocation of resources
> > is not computational costs, but mental costs:
> >
> > http://szabo.best.vwh.net/micropayments.html
> >
> > Mental accounting (or mental transaction) costs arise whenever
> > Grandma (or anybody else) is asked to make a security or economic
> > decision.  Mental costs pose a severe limit on how fined-grained such
> > decisions can be, and thus on how fine-grained it is useful
> > for the computational mechanisms (such as capabilities for
> > security or tokens for payment) to operate.
>
>I agree, but I don't think about money when I turn on the light or
>get a drink <of water from the tap>.
>Both of these run a meter which turns directly into a bill.

I think the above summarizes quite tersely the position I was trying
to layout for "accounts" (meters) vs. direct payments (cash,
micropayments).  It does seem to me that if suitably micro
and inexpensive payments are available then one could implement
something like a meter with them, but I believe that something
like such a meter facility is needed (valuable) at some level.
That level I believe would most suitably include an object level
permission for the "meter" and so would be a capability (Capability
accounting, the subject of this thread).

If others believe that such a mechanism isn't needed, then I'd be
quite interested to hear how they imagine charging, for example,
for processor time in a time sharing system.  Naturally I'm content
with anything that works effectively.  At the moment I just can't
see how such a mechanism would work without the equivalent of
a "meter".

If we do have a consensus that something like a meter is needed,
then I believe the issue of how to deal with communicating some
sort of "sub meter" to services and how to do reporting for such
charged services and ultimately to "cut off" charging for any
such services naturally arises along with it's relationship to
the "lost object" problem.

Perhaps I should mention that my goal with this discussion is
entirely practical.  I don't see how services like Wideword are
ever going to flourish (and with them network capabilities and
a wedge into a POLA IT infrastructure) if they can't manage
(control, limit, account for) their resources.  Naturally for all
the reasons that capabilities are a good idea, the management
of resources needs to be done with capabilities - hence this
discussion.  I would like to have a consensus model readily
available for anybody at the point of implementing such network
services that are managed via network capabilities (e.g. me ;-).

--Jed http://www.webstart.com/jed/