F21 Self Contained Change: Security Policy In The Installer

[Matthew Garrett]

On Fri, Mar 14, 2014 at 06:24:36PM -0400, Eric H. Christensen wrote:
> On Fri, Mar 14, 2014 at 08:01:53PM +0000, Matthew Garrett wrote:
> > If an incorrect choice means that the software the user wants to run 
> > won't run, that's going to be a problem for the user. And we presumably 
> > expect that some software won't run, because otherwise we'd be enabling 
> > that security feature by default? A user who accidentally installs a 
> > profile that enables FIPS compliance is going to have a bad time, for 
> > instance.
> 
> No, that's not exactly it.  I've pointed out reasons why defaults 
> usually suck (security-wise).  I've yet to see a hardened system make 
> software fail.  I'd love some examples of your concerns.  I also don't 
> understand why FIPS compliance will make a user have a bad time since 
> I've been on systems that were fully FIPS compliant and didn't have 
> any problems.

You don't think it would upset users to have their kernel panic if 
they accidentally tried to load an inappropriately signed module? What 
happens if I ssh to a server that doesn't implement any of the 
FIPS-approved algorithms? Why is Firefox suddenly asking for a password 
before I can visit https sites? Why won't Firefox speak https to a bunch 
of sites?

-- 
Matthew Garrett | mjg59 at srcf.ucam.org