On 05/04/13 22:27, Ariel Ben-Yehuda wrote:
> Hi,
>
> I could probably do #703 (I am familiarizing myself with that area of the
> code) and maybe #441.
>
"Could probably"? Hey, I "could" work on any of those tickets, the
question is whether I _will_ work on any of them! There is a big
difference between "could" and "will" ;)
Anyway, because Andrew Sorenson has already done some work on this, and
declared to work on this in the coming weeks, so let's leave this ticket
to him.
> I think that for #441, following the "don't mix explosives and detonators"
> principle, we should encrypt the files under a "null key" to prevent
> accidents and then make qvm-open-with-dvm decrypt with said "null key".
>
> The "null key" should be publicly known (even the zero key will suffice)
> but we use it to prevent accidents (say, running `evince *.pdf' or creating
> thumbnails on the filesystem).
>
> This has the problem of preventing Bittorrent, but one shouldn't put it on
> a trusted domain anyway...
>
Let's move this discussion to a new thread for better clarity and easier
referencing. Ariel, can write a new message describing the overall
approach to this task? What you mentioned above is just one little
element, while many other remain to be answered. E.g. how would you
"hook" the mime handlers in the first place? How would you remember the
user choices (Open natively, Open in DispVM, Open in another VM, Send to
converter) for each file?
I don't quite like the approach with null keys, or any approach with
requires modification of the original files that could prevent opening
of them on other systems. We don't want to lock people to Qubes OS,
we're not Apple :)
So, anyway, please give this a thought, and please describe your plan in
a new message in a new thread that could be nicely linked in from the
ticker.
Thanks.
joanna.