wosym
commented
2 years ago FYI: first link is not publicly accessible. (Not even with an account)
Open benpicco opened 2 years ago
wosym
commented
2 years ago FYI: first link is not publicly accessible. (Not even with an account)
The recent forum thread about what we would do if we had significantly more engineering resources sparked many good ideas, with many sensible suggestions to improve the existing code base.
There are always some people who are doing a research project / thesis related to RIOT, there cleanups and refactoring are not very exciting proposals. So here are some juicy topics that I could think of, which, while ambitious, could also be good way to get into RIOT development.
Feel free to edit this post if you can come up with more ideas / details.
'Fuzzy' Timers
Add an API to schedule a wake up with a min / max deadline. RIOT should try to coalesce timeouts that fall into the same interval to reduce overall wake-up events.
Linux has
usleep_range()for this.Multiprocessor Support
No SMP as MCUs with multiple cores are not always symmetric (e.g. only the network core can access the radio registers). Add a mechanism to assign threads to a CPU core and bring up the additional cores.
Secure wireless networks
IEEE 802.15.4 MAC layer security exists, but is only the first building block for a secure network. It relies on a shared key that needs to be exchanged on network join and that should be exchanged periodically.
A certificate mechanism for nodes to authenticate themselves to the coordinator could be imagined.
6LoCAN
Just like 6LoWPAN, but with CAN bus instead of IEEE 802.15.4. Thesis, RFC draft
Porting RIOT to a new MCU family / architecture
This is always an option as there is still plenty of hardware not supported by RIOT. There is a forum thread with some suggestions, but there are always new MCUs being released and many vendors not supported by RIOT at all.
Integrate LPWAN optimizations in
gnrc_tcpGNRC has a very basic TCP implementation. We had a talk (video) about optimising TCP for low power wireless networks, but that was based on BSD's TCP implementation. Having this properly integrated into RIOT would be a huge benefit for many applications.