Re: Object Capability model for IoT

On Tue, June 17, 2014 18:50, David Barbour wrote:
> On Mon, Jun 16, 2014 at 4:28 AM, Tim Coote <[email protected]>
> wrote:
>
>> The aspect that I’m struggling most with is how to ensure that
>> capabilities don’t just become strings
>
>
> Using strings for capabilities is fine, so long as they aren't
> (cryptographically) forgeable. Sometimes these are called 'password
> capabilities' or 'sparse capabilities'. I tend to favor meaningful strings
> that are signed with HMAC or PKI, or simply encrypted strings, to
> represent
> capabilities.
>
> Avoiding large random strings helps avoid stateful table maintenance or
> distributed garbage collection. Using encrypted strings as capabilities
> also admits an interesting possibility of leveraging homomorphic
> encryption, i.e. some ability to operate upon a capability (e.g. to
> attenuate it, block facets) without cross-network communication.

Could you elaborate on when/why using random strings would induce the need
for distributed garbage collection?