FOSDEM 2016

It the beginning of the year and, surprise, FOSDEM happened :-) This year I even managed to get to see some talks and to meet people! Still not as many as I would have liked, but I’m getting there…

Lenny talked about systemd and what is going to be added in the near future. Among many things, he made DNSSEC stand out. I not sure yet whether I like it or not. One the one hand, you might get more confidence in your DNS results. Although, as he said, the benefits are small as authentication of your bank happens on a different layer.

Giovanni talked about the importance of FOSS in the surveillance era. He began by mentioning that France declared the state of emergency after the Paris attacks. That, however, is not in line with democratic thinking, he said. It’s a tool from a few dozens of years ago, he said. With that emergency state, the government tries to weaken encryption and to ban any technology that may be used by so called terrorists. That may very well include black Seat cars like the ones used by the Paris attackers. But you cannot ban simple tools like that, he said. He said that we should make our tools much more accessible by using standard FLOSS licenses. He alluded to OpenSSL’s weird license being the culprit that caused Heartbleed not to have been found earlier. He also urged the audience to develop simpler and better tools. He complained about GnuPG being too cumbersome to use. I think the talk was a mixed bag of topics and got lost over the many topics at hand. Anyway, he concluded with an interesting interpretation of Franklin’s quote: If you sacrifice software freedom for security you deserve neither. I fully agree.

In a terrible Frenglish, Ludovic presented on Python’s async and await keywords. He said you must not confuse asynchronous and parallel execution. With asynchronous execution, all tasks are started but only one task finishes at a time. With parallel execution, however, tasks can also finish at the same time. I don’t know yet whether that description convinces me. Anyway, you should use async, he said, when dealing with sending or receiving data over a (mobile) network. Compared to (p)threads, you work cooperatively on the scheduling as opposed to preemptive scheduling (compare time.sleep vs. asyncio.sleep).

Aleksander was talking on the Tizen security model. I knew that they were using SMACK, but they also use a classic DAC system by simply separating users. Cynara is the new kid on the block. It is a userspace privilege checker. A service, like GPS, if accessed via some form of RPC, sends the credentials it received from the client to Cynara which then makes a decision as to whether access is allowed or not. So it seems to be an “inside out” broker. Instead of having something like a reference monitor which dispatches requests to a server only if you are allowed to, the server needs to check itself. He went on talking about how applications integrate with Cynara, like where to store files and how to label them. The credentials which are passed around are a SMACK label to identify the application. The user id which runs the application and privilege which represents the requested privilege. I suppose that the Cynara system only makes sense once you can safely identify an application which, I think, you can only do properly when you are using something like SMACK to assign label during installation.

Daniel was then talking about his USBGuard project. It’s basically a firewall for USB devices. I found that particularly interesting, because I have a history with USB security and I do know that random USB devices pose a problem. We are also working on integrating USB blocking capabilities with GNOME, so I was keen on meeting Daniel. He presented his program, what it does, and how to use it. I think it’s a good initiative and we should certainly continue exploring the realm of blocking USB devices. It’s unfortunate, though, that he has made some weird technological choices like using C++ or a weird IPC system. If it was using D-Bus then we could make use of it easily :-/ The talk was actually followed by Krzyzstof who I reported on last time, who built USB devices in software. As I always wanted to do that, I approached him and complained about my problems doing so ;-)

Chris from wolfSSL explained how they do testing for their TLS implementation. wolfSSL is 10 years old and secures over 1 billion endpoints, he said. Most interestingly, they have interoperability testing with other TLS implementations. He said they want to be the most well tested TLS library available which I think is a very good goal! He was a very good speaker and I really enjoyed learning about their different testing strategies.

I didn’t really follow what Pam was talking about implicit trademark and patent licenses. But it seems to be an open question whether patents and trademarks are treated similarly when it comes to granting someone the right to use “the software”. But I didn’t really understand why it would be a question, because I haven’t heard about a case in which it was argued that the right on the name of the software had also been transferred. But then again, I am not a lawyer and I don’t want to become one…

Jeremiah referred on safety-critical FOSS. Safety critical, he said, was functional safety which means that your device must limp back home at a lower gear if anything goes wrong. He mentioned several standards like IEC 61508, ISO 26262, and others. Some of these standards define “Safety Integrity Levels” which define how likely risks are. Some GNU/Linux systems have gone through that certification process, he said. But I didn’t really understand what copylefted software has to do with it. The automotive industry seems to be an entirely different animal…

If you’ve missed this year’s FOSDEM, you may want to have a look at the recordings. It’s not VoCCC type quality like with the CCCongress, but still good. Also, you can look forward to next year’s FOSDEM! Brussels is nice, although they could improve the weather ;-) See you next year!

3 thoughts on “FOSDEM 2016”

  1. Cynara sounds exactly like polkit — I haven’t compared them, but I’m wondering why they’re not using polkit in Tizen.

    For completeness, the way this thing is typically done on desktop Linux is (when using D-Bus for RPC) to get the peer’s security label using GetConnectionCredentials, then to pass those to polkit to make a policy decision on.

    1. I think I asked how it compares to PolKit. But I’ve forgotten what he said… It may have been caught on the recording.

Comments are closed.

Leave a Reply

Your email address will not be published. Required fields are marked *