Tracking: Userland UDP Interface Milestone

[hudson-ayers]

This issue dictates what tasks must be complete for the Userland UDP Interface milestone to be considered completed. Please add to this as needed.

• Transition the Imix initialization for UDP/6LoWPAN to use the component interface
• Clean up configuration so that all networking stack configuration happens either in main.rs or via userspace configuration calls. This includes creating some configuration file from which the src MAC, src PAN, RF channel, and src IP addresses are pulled. This also includes creating an interface so that the IP addresses of local interfaces can be accessed by userland apps via the udp driver.
• Enforce port bindings to prevent one application from listening to messages intended for another.
• Add additional return methods to existing system calls to reflect failures due to unavailable ports
• Enforce that apps can also only send on ports they are bound to
• Reach agreement that the interface between userland and the kernel is finalized
• Have someone other than Hudson use the finalized interface (hopefully while working on the Sensys tutorial)
  • Distribute imixes to other developers for testing.
    • Ensure we have enough imixes for sensys tutorial.
• Test behavior when multiple userland apps use the driver simultaneously
• Resolve issues sending/receiving multi-fragment udp messages
• Either allow userland apps to send 1280 byte packets, or a driver call to learn the max size packet that can be sent.

[bradjc]

Reach agreement that the interface between userland and the kernel is finalized

With plans of doing a release in the next couple weeks, it seems unlikely (or at least I think it's unlikely? We still haven't stabilized a GPIO interface...) that we will come to a consensus on this, and we haven't gotten too many hours of use if you will on the interface either. Does it make sense to move this to a new milestone? I think userland UDP is more than enough of a milestone for the release, without having to say the syscall interface is set.

[hudson-ayers]

Yes, I agree with that, especially given the extent to which we (@phil-levis in particular) have stressed not rushing stabilizing interfaces

[bradjc]

I think the only thing left in this issue is testing, which we need to do for the release anyway.