Autonomous smartphone apps: self-compilation, mutation, and viral spreading | Hacker News Reader

Autonomous smartphone apps: self-compilation, mutation, and viral spreading

arxiv.org

46 pointsingve12 comments

I'm not that impressed by this.

They talk a lot about "mesh networks" in the "Related Works" section in the introduction. Their work is in fact not related to any kind of mesh network. The words "mesh network" are a very loose collection of concepts, and are used to describe certain classes of IP routing protocols, large volunteer-run ISPs like guifi.net and Freifunk, and p2p messaging apps like Firechat. The authors do not seem to understand the distinction, and their analysis of this "related work" does not rise above the level of "wow, cool decentralized stuff".

On to the actual work being presented, it looks to me like the researchers just put an Android compiler into an Android app (Maybe this is a big accomplishment? I don't know much about Android development). In any case, it has an interface to select a few superficial options for the app, like the icon, and then it will compile a new copy.

Using NFC, the compiled app can be transferred to another phone. This has nothing to do with mesh networks in any sense of the word. They also present an analysis of compile times on various phones, which top out at 330 seconds. Useful information?

In any case, it's cool that they put a compiler in an app, but this paper feels like a massive edifice of fluff built up around a cool weekend hack. I hope that they can pass their class or get their certificate or degree or whatever they're gunning for here.

I agree with your analysis. I truly hope that people motivated to make a difference by enabling distributed resistance to censorship of centralized communications infrastructure can see the wisdom in investing time to develop the wifi mesh capability for FirefoxOS or Android. Once a major platform has this functionality, it will spread. I wrote a fairly long paper about ideas in this space over here: https://bug945047.bugzilla.mozilla.org/attachment.cgi?id=840...

I haven't read your paper yet but have you looked at Serval?

Paper author here... In our main project we implemented an onion routing stack compliant with the Tor protocol specification. Key issue is we avoid using any central server, like directory servers. It uses UDP (not TCP like Tor itself), runs on Android, fully functional, but not mature yet. See our overview talk, http://wan.poly.edu/p2p2015/keynote.html

> The authors do not seem to understand the distinction

We are very much aware of the difference between ad-hoc and NFC transfers. Note that for dissent and anti-government protest you most likely need something else: Delay/Disruption Tolerant Networking (DTN), see these IETF pages, http://ietf.org/dyn/wg/charter/dtn-charter

> I don't know much about Android development

This is quite non-trivial.

> I haven't read your paper yet but have you looked at Serval?

Yes, we are in contact with them. We invited the Serval people to one of the IETF events I organised some time ago. Sadly they are a bit underfunded. http://www.internetsociety.org/articles/moving-toward-censor...

Sorry for the harshness of my tone. There's unfortunately a lot of people draping their tech in the mantle of "mesh" without actually introducing any innovation. I had taken this as such a project mostly because I saw no mention of routing protocols. I will take a look at your main work.

I agree that, theoretically speaking, it's not all that impressive, but in practice, to be able to create and/or modify apps for an Android device using an Android device is nice since it gives more visibility to the idea that Android devices are also general-purpose computers. In fact, one of the FSF's complaints againt Android is that it's not self-hosting:

http://www.gnu.org/philosophy/android-and-users-freedom.en.h...

The github repo[0] for this was posted on HN a few days ago and garnered a bit of discussion[1].

[0] https://github.com/Tribler/self-compile-Android

[1] https://news.ycombinator.com/item?id=10529021

wrong link to the discussion

correct prior news.ycombinator link: https://news.ycombinator.com/item?id=10517936

We do seem to be building in a lot of risk to our infrastructure at the moment from network connections to software delivery and introducing many single points of failure. It's good for a company that wants to control how you connect to the Internet, what you consume and to generate revenue from that but it's not great in any situation where those single points of failure...fail. We do need to consider alternative approaches which include peer delivery and mesh networking. We have the technology to do that now.

Congrats on inventing skynet.

Self compiling is not self-sentient.

Reply on news.ycombinator.com