[prev in list] [next in list] [prev in thread] [next in thread] 

List:       cap-talk
Subject:    Re: [cap-talk] building a model
From:       John Carlson <john.carlson3 () sbcglobal ! net>
Date:       2005-05-16 0:33:54
Message-ID: 4287EA72.2090007 () sbcglobal ! net
[Download RAW message or body]

David Hopwood wrote:

> John Carlson wrote:
>
>> One of my questions would be, for capabilities as keys versus
>> object capabilities is:  Which is more economically feasible?
>> Are the efforts required for implementing both equivalent?
>>
>> I think that one also needs to consider the case where keys
>> are subjects (serialized perhaps), not just resources.  This
>> may aleviate some of the concerns about keys.
>>
>> The nice thing about keys is that they can be sent through
>> email, over the web etc., copied and pasted.   Do we have
>> object-oriented email yet?   Capability secure clipboards?
>
>
> You seem to be assuming that capabilities-as-data and
> capabilities-as-object-references are mutually exclusive.
>
Perhaps.  I was just looking at the capmyth reference, and
trying to figure out whether I prefered the as-data versus
the as-reference ideas.

> Many systems provide both, e.g. E provides local and remote
> object references (with different failure semantics), plus
> "sturdy refs" which are just data. The references are more
> efficient and act as object capabilities (which have some
> technical advantages related to confinement), but can only
> be used either locally or in E's CapTP/VatTP protocols.
> Sturdy refs, OTOH, can be sent in email and other non-capability
> protocols. The cost of supporting both is not a significant
> problem.

Thank you, that was the answer I was looking for.

John

_______________________________________________
cap-talk mailing list
cap-talk@mail.eros-os.org
http://www.eros-os.org/mailman/listinfo/cap-talk
[prev in list] [next in list] [prev in thread] [next in thread]