Dear all,

draft-pjs-6lo-can-01 is posted. I have a short slot at the end of Wednesday's
6lo session — last on the agenda, so it may not survive the earlier
discussions. This message is therefore the substantial version: what the
document is for, where it came from, and what I would like to discuss, here
on the list.

  https://datatracker.ietf.org/doc/draft-pjs-6lo-can/
  
https://author-tools.ietf.org/iddiff?url1=draft-pjs-6lo-can-00&url2=draft-pjs-6lo-can-01

GOALS

The case is equipment where the CAN bus is part of the certified product and
cannot be replaced, while the devices on it are now expected to be reachable
over IP for monitoring, configuration and software update. My coworkers'  is in
service: the elevator management platform M2M Tech built for LH, the Korean
national housing authority, running CAN 2.0B with 29-bit extended
identifiers at 50 kbit/s. CANopen machinery and building automation have the
same shape. Today this is bridged by a gateway translating a vendor-specific
CAN protocol into IP, one per combination of equipment and management
system.

The draft defines a 14-bit node address mapped into the 29-bit CAN frame
identifier, a link-layer duplicate address detection so those addresses can
be assigned at random, ISO-TP for segmentation, and an Ethernet border
translator. A segment is a single link, so RFC 4861 Neighbor Discovery is
used as it is.

HISTORY

This continues draft-wachter-6lo-can, presented by Alexander Wachter at
IETF 106 and reviewed by Carles on this list in June 2020 — 35 comments that
never got a revision in reply. From the IETF 106 minutes:

  Pascal: seems SCHC would make sense for the 6loCAN case.
  Carles: is ISO-TP the native fragmentation mechanism for 6LoCAN?

-01 is that reply. Compression moves from LOWPAN_IPHC to SCHC, while ISO-TP
keeps providing fragmentation and the document now says why. CAN FD is
required, because eight octets minus the ISO-TP header leaves nothing for a
compressed header. Remote frames are gone from the duplicate address
detection. RFC 8065(privacy) is taken into account. The slides, the minutes, 
the 2020
review and a disposition of all 35 comments are here:

  https://github.com/fnumber/6lo-can/tree/main/history

DISCUSSION ITEMS

1. Is the split right — compression from SCHC, fragmentation from ISO-TP?

2. -01 fixes the SCHC Rule ID at 8 bits and leaves the rules themselves to
   the installation. That gives a working network but not an interoperable
   one. Does an example rule set belong in an appendix?

3. Coexistence. 6LoCAN claims the whole 29-bit identifier space and leaves
   the 11-bit identifiers free. That covers CANopen but not J1939 — nor the
   deployment above, which already uses 29-bit extended identifiers for its
   own addressing. As written, 6LoCAN would collide with the very system it
   is meant for. The document needs a coexistence section.

4. What would the group want to see before this could be considered for
   adoption? I expect an implementation. I do not have one yet, but I do
   have the deployment it would go into, and access to it; I would be glad
   of company in building it.

Review of the addressing scheme and of the duplicate address detection is
the most useful thing right now — those are the hardest parts to change once
there are implementations.

Alexander is a co-author and will join the session on Wednesday.

Thank you,

Jungsoo Park
Electronics and Telecommunications Research Institute (ETRI)
[email protected]

_______________________________________________
6lo mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to