On Mar 16, 2012, at 10:49 AM, Justin Lebar wrote:

>> Yes, clearly OWA was not designed with Gaia apps in mind.   To be blunt, my 
>> opinion at this point is that a model with no code authentication or 
>> controls on importing code over plaintext channels, is insufficient for a 
>> privileged application like Gaia.  It would leave Gaia apps open to the most 
>> trivial MITM attacks.
>>  Lucas.
> 
> I understand the bit about code authentication.  If the web server
> gets hacked, we're screwed.
> 
> But surely well-functioning Gaia code would only load code over HTTPS,
> so I don't understand where this MITM attack comes from.

Sure, I think if we believe that all Gaia apps should be loaded over HTTPS then 
we should just enforce that.  Having worked through this problem before, my 
faith in people voluntarily using HTTPS when necessary is really low. 

That doesn't address the system risk issue, which is something I'd like to try 
to figure out.  Jim Straus has a proposal in the other thread of including 
hashes in the manifest, though that would require signing the manifest itself 
in turn with something that chains to a trusted root… blah blah… blah.  
Otherwise you're still just relying on HTTPS.
  Lucas.
_______________________________________________
dev-security mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-security

Reply via email to