On to, 2015-07-09 at 11:15 +0200, Marcel Holtmann wrote:
> Hi Jussi,
> 
> >>> Now that we're talking about the online check... I've talked to people
> >>> who considered this behaviour "calling home" and thought it
> >>> unreasonable that
> >>>   a) it's not possible to prevent the online check from happening via
> >>> configuration and
> >> 
> >> 
> >> Why would you prevent it to happen? This check is something extra anyway,
> >> being "ready" means you are connected. In a way, people should not care too
> >> much not seeing  there service being set to "online", since such check 
> >> can't
> >> be bullet-proof.
> > 
> > It's not about connman functionality at all. It's (as an example)
> > about people building a super secret embedded product demo on top of
> > Yocto suddenly realizing that their device is connecting to a web
> > server they don't control or even know about (aka "who is this Marcel
> > Holtmann why is our IOT Fridge fetching web pages from him?")
> 
> if this super secret fridge is connected to the Internet and can actually 
> reach it, then it is no longer super secret. If you would be really worried, 
> then you would have it locked up in a lab with no access to anything.
> 
> And even if it would be calling the ConnMan servers, nobody in the world 
> could tell super secret fridge apart from someone sitting next to it using 
> ConnMan on Yocto on a Minnowboard or its laptop.
> 
> I can not repeat this enough. This whole think is designed with full 
> anonymity in mind. We are building our own HTTP request for exactly that 
> reason. No headers can sneak in. No unwanted meta data can leak. ConnMan 
> ships its own HTTP client for a reason.
> 
> Think about this for a second. You can leak more information by using a 
> distro libcurl by accident that includes some meta headers. If ConnMan's 
> online check is your concern, then you do not understand privacy at all.
> 
> > FWIW, I totally understand the point Patrik makes in the previous 
> > discussion:
> >> If the URL is configurable, upstream
> >> does not have the means to fix online related bug reports as we'd be
> >> unable to confirm the online checking service to work properly in the
> >> first place. In the worst case even the URL accessed is not known, not
> >> even to the person submitting the bug report.
> > 
> > This is a compelling argument as well -- I'm not in favor of making
> > this too easy. I do think there are legitimate use cases where people
> > do not want to rely on Marcel (or Intel) handling their online checks.
> > 
> > Since the required patches are not big (as shown by Pasi, thanks!)
> > just adding clear documentation to Yocto about the online check
> > Connman makes and instructions on how to modify it may be a sufficient
> > alternative.
> 
> If someone wants to add documentation on what ConnMan is doing, then I am all 
> for it. More documentation is always good. I am in favor of full transparency.

There is already documentation about online check in ConnMan README
file. If that is not enough, then patches are welcome.


> 
> I am however not in favor of giving such an option. If someone wants to shoot 
> themselves into the foot, then they can pick up the gun by themselves. I do 
> not see a good reason why would make this easy.
> 
> And honestly I prefer them carrying an extra patch changing the defines in 
> the source code. That means they have to carry that extra patch. So every 
> time it breaks, they get a nice reminder that something might have changed. 
> So they have a change in catching it. The config file is to easy to forget.
> 
> Regards
> 
> Marcel


Cheers,
Jukka


_______________________________________________
connman mailing list
[email protected]
https://lists.connman.net/mailman/listinfo/connman

Reply via email to