Today, I did a presentation about the SmolPhone project to the Eco ICT school, and it went smoothly.

It's an update over the presentation I did back in February in Grenoble. The presentation flow is smoother, the motivation hopefully more explicit: designing a phone is a mean, not a goal. We really want to understand how modern computing stacks are built and try to debloat them. Building a phone is a research action toward our goal, not the goal itself.

The take-home message is that working on such topic is not a revolutionary change to my research agenda actually, but merely an evolution. After 10 years on performance modeling of distributed infrastructures, and then 10 years on correction of distributed infrastructure, I'm looking forward on post-growth infrastructures. I think they need less bloat (simplicity over modularity and other things) and they need to cope with a fluctuating world (robustness over performance; interactions over hierarchies). But once you've fixed these goals and criteria (or others), then it's a good old research agenda that we can work on as usual. It's not business as usual either because the goal is not taken as granted from the Big Fish in the pond (in the mobile world, that would be the GAFAM). I know that tech and science are not neutral, and I thought about the world I'd like to help foster before choosing these goals.

(pdf slides)

We had an interesting discussion about one of the crazy/stupid idea I proposed: sending data over the voice that is encoded in GSM. My preliminary formulation (IP-over-VoIP) was funny but very wrong. The goal is to use the fact that I do not pay for voice connection while a voice call over GSM consumes less energy than maintaining a 4G connection, and try to send data over the voice channel of GSM. The analogic encoding of your message must be robust to the signal simplification operated by your codec. The rate will be bad, but I'm still wondering what we could get this way.

I was told that if I need an old phone at home to host the services that are mandatory to the smolphone, then maybe I could keep that old phone in my pocket and save the effort of designing yet another device that will end in the landfill at some point. That's a good point. I said that the smolphone terminal is fully open and easier to understand, evolve and fix. But I need to think further about why I still think that it's better to have two devices.

A friend reacted to this blog post, criticizing this idea of junkyard computing to help self-hosting. From his (extensive) experience in self-hosting in the FFDN association, the good scale is when someone hosts the services for at least a dozen of people, not only its own. The reason is that it's simply impossible to simplify self-hosting beyond a given point, and you want a friend with good technical skills when something goes wrong (e.g. when you are under a malicious attack). That's a very appealing alternative to my half-backed idea. We could dream of communities of users working at the local scale to host the services they need. That'd be in line with the whole idea of low techs, for sure.

I was asked why I do not want to explore the possibility of fully optimizing Linux phones (the ones without Android, such as a good old Galaxy S5 running under PostmarketOS) to see how far we can go. I personally think that it's a dead end because Linux is becoming a monster of genericity. It starts to feel quite bloated on my laptop. As the people trying to integrate Rust into the kernel painfully experience, getting Linux to evolve takes an inordinate amount of energy and work. I prefer to experiment on a niche and explore new possibilities without the burden of getting the whole world to agree with each of my attempts.

I was told that the goals of one week for the battery life and one decade for the device life are not ambitious enough and that's probably true. I guess that when we have the device for real, we will try to push the limit forward. But for the time being, one week and one decade are simple to remember and that's all what matters for now.

The day before, this audience had a presentation of Stéphane Crozat about a fantasy phone that would last your entire life: LifePhone. It's an exercise of thinking where the participants discuss of what features they'd love to see on a future phone. Even if the SmolPhone is still far from reality, I think that both projects are complementary. Someone asked me about the most controversial point (IMHO) of the lifephone: one of the design decisions debated by the participants is whether the LifePhone should be the same for everyone (yielding easy maintenance, slow evolutions among other advantages), or whether each device should be personnally crafted (yielding some affective attachment of the user to the device, possibly down to addiction).

It's a very good question, very controversial. I do have a strong opinion here: I foster a specific connection between the user and the device, because I think that these devices are too fragile to be badly handled. Most of them get replaced after falling to the floor, so I really think that users should take more care of these fragile devices. I build upon a very good paper that was published at ICT4S last June, but unfortunately, the pdf is not to be found on the internet at the moment. I think that the addiction aspect can be addressed otherwise, e.g. through social medias fighting against this addiction (like Mastodon) instead of building on it (like Twitter). Also, I think that removing the video features will be very effective against the addiction.

I was asked about the power consumption of a voice call on a 5G modem, but this is not something we explored at all. Actually, we should probably redo all measurements in a common framework to make them comparable. And this person was right, we should get some 5G modems too in addition to the 4G LTE Cat 1 ones that we currently target. It seems that the low power mechanism invented in Cat 1 such as Power Saving Mode (PSM) and extended delayed reception (eDRX) made their way to 5G too. Adding a 5G modem would not be in line with the goal of debloating the world, but at least we need to measure it.

I'm eagerly waiting for the two engineers to arrive and start working on that project. I'm eager to see the first technical difficulties cleared and get a better vision of the tradeoffs at stake. Victorien should be there in December, and the person working on the hardware should be there in January or February. I'm so excited!