Re: [core] CoAP over Serial URLs

[Carsten Bormann <[e-mail]>]

Hi Tobias,


> On 14 Dec 2016, at 19:30, Tobias Kaupat <[e-mail]> wrote:
> 
> Hi all,
> I'm implementing CoAP over a serial RS232 connection

Interesting — we had a brief discussion about this at the T2TRG meeting in Ludwigsburg, and I’d sure like to know more about this.

> and wondering about the URL format.

When we started to think about the bluetooth URI format (draft-bormann-t2trg-ble-uri) we ran into a similar discussion.  A URI for a “local interconnect” isn’t really a “U”RI, as its meaning is dependent on that local context.  There is precedence in the file:// URI, but there is still some uneasiness around.  (There are also authority-based URIs with link-local addresses; similar problem.)

> For CoAP over SMS they use a separate scheme "coap+sms" (see: https://www.ietf.org/archive/id/draft-becker-core-coap-sms-gprs-05.txt) so I just took "coap+rs232" for my case.

Are the voltage levels on the serial line (which is what “RS232" is about) really important?
Do you need a different URI for TTL-level UARTs?  Maybe more like coap+serial or coap+uart.

How do you decide the bit rate?  The CoAP over Serial specification could align on a strong default (say, 115200 bit/s as in 6lobac), but there may occasionally be a need to deviate even from such a strong default.

> So far the only question I have is: Is there already some other URL scheme defined that I'm not aware of?

Not for serial.

> The actual question is regarding the Host. A serial connection is usually a file handle. For Windows something like "COM3" is just fine. On Linux based systems I have to connect to "/dev/ttyS3" which gets more problematic when encoding in URLs. What is the best way to deal with the slash characters?
> 
> 
> https://tools.ietf.org/html/rfc3986#page-21 restricts percent encoding of ASCII characters in the host part but allows system specific host lookups. Thats why I came up with an Implementation that implicitly adds the prefix "/dev/" to the host part of the URL before opening the connection on Linux systems.

That is probably the right approach.  Again, compare the link-local addresses which would need a % character (which need to percent-encoded) — endless discussion.

I’m definitely interested in also nailing down the data format over serial.  Let’s discuss this tomorrow...

Grüße, Carsten