Re: Fwd: Re: [Cryptography] Encryption opinion

On Tue, September 2, 2014 18:31, [email protected] wrote:

>> > * Pages may grant one another access to local storage by sharing
>> > capabilities
>>
>> Sounds promising, If only I had some more spare time the above two
>> points
>> would be a great use-case for a ChromeOS fork that uses my (slowly)
>> ongoing work on RumpelstiltskinFs and friends ;-) Unfortunately my
>> spare-time night already be to small at the moment to even finish
>> RumpelstiltskinFs itself any time soon :-(
>
> If only we all had time. I'll look at RumpelstiltsinkFs ... never heard of
> that before.

My apologies. I recently split up what was previously MinorFS-2 into a set
of different projects:

https://github.com/pibara/MinorFs2/blob/master/README.md

RumpelstiltskinFS is to be an extended version of MinorFS::capfs, a
sparse-cap based overlay file-system with support for decomposition and
read-only attenuation.

For some more context:

http://ohm2013.capibara.com/
http://pibara.github.io/Rumpeltreepp/

I think if a multi-process browser like chrome could somehow communicate
active-page properties from a primary process about child processes, the
file-system could expose only those chunks of page private storage the
specific process owns trough its currently active page. This would mean
that even if the browser had a vulnerability, no private storage could be
exposed by an exploit of such a vulnerability. If I ever complete all of
the current MinorFS derived projects, I would love to have a look at the
possibilities for something like that.

> The key here -- and that's what led me to my second line of thinking
> upthread, the one that I withdrew -- is that, under these circumstances,
> the Web browser becomes a powerbox, granting not only local authorities
> (the camera, etc.) but also mediating and keeping track of the grants
> between one page and another, a la the Web Introducer.
>
> In this new same-origin-free world, we'd need to come up with some way to
> assign responsibility for capabilities granted -- "to whom" where they
> granted? What set of pages | objects | UI affordances should we now assume
> can wield capability C?
>
> And if these pages embed third-party objects of their own, they are then
> acting as nested powerboxes. How can we best represent that, short of just
> nested visual boxes within boxes within boxes?
>
> Ihab
>