Re: Object Capability model for IoT

On Mon, Jun 23, 2014 at 10:09 AM, Karp, Alan H <[email protected]> wrote:

>  I don’t see why the mapping table is a big issue.  If it grows without
> bound, that means the number of objects grows without bound, too.  Even
> with meaningful strings, you have the same distributed garbage collection
> problem.
>

It is true that "the number of objects grows without bounds" either way.
But you're wrong about the implications on GC.

Consider: when object information is encoded directly in the string, you
can 'drop' that information by the simple (and entirely local) act of
dropping the string. However, when object information is encoded in an
extrinsic table, however, the information sticks around long after you drop
the capability, and so you need a distributed garbage collection process to
eventually remove the information from the system. Fundamentally, it is
these 'mapping tables' that are GC'd, not 'objects'.

Random strings, by nature of being random, do not encode information, and
so information must always be encoded extrinsically, using a mapping table
or equivalent. Meaningful strings, by comparison, can often encode
stateless objects or functions directly, and can further encode an
unbounded set of facets and compositions and such that guard access to a
constrained subset of stateful objects or resources.

_______________________________________________
cap-talk mailing list
[email protected]
http://www.eros-os.org/mailman/listinfo/cap-talk