Implementation report for aiocoap

[chrysn]

Beyond the things that gave me trouble, one aspect not covered in the document are port numbers. They don't really have a place in the transport, but setting them in a Uri-Port option might be practical, especially for when an MCU has more UARTs it fans out to, as that'd allow the main entry point to do simple port based reverse proxying (with barely any real need to do more than forwarding messages, as long as it does not issue requests itself downstream), and then they need support in the URI. This might be easier though when going with the more generic coap:// scheme as per my transport-indication recommendations.

[cabo]

Is it the hostname or the port that you would use for locally selecting UART lines?

[chrysn]

hostname or port

Either would be viable, and both probably need some attention. I just think that it's a bit easier for a very constrained "border proxy" to manage its back-ends by numbers (counting up, discarding leading zeros or even using over-long CoAP uint options) than coming up with names (base-32'ing a count or otherwise filling the allowed byte values of a UTF-8 string). And especially when people are asking questions about using this on a shared RS485 bus, where hostnames would be the more natural places for participant identifiers.