> >> Would you prefer to list ECN and DSCP separately and in the order >> that they appear in a LOWPAN_IPHC encoded header? For example: >> >> 00: ECN + DSCP + 4-bit Pad + Flow Label (4 bytes) >> 01: ECN + 2-bit Pad + Flow Label (3 bytes) >> 10: ECN + DSCP (1 byte) >> 11: Version, Traffic Class, and Flow Label are compressed. > >Great. Concise and precise. I'd maybe add: > >> 00: ECN + DSCP + 4-bit Pad + Flow Label (4 bytes) >> 01: ECN + 2-bit Pad + Flow Label (3 bytes), DSCP is elided >> 10: ECN + DSCP (1 byte), Flow Label is elided > >> 11: (0 bytes), Traffic Class and Flow Label are elided. >> All elided fields are decompressed as zero, except that Version is >> always elided and decompressed as 6. >> As usual, bit fields are concatenated by filling the most >> significant bits first. > > >> >> There is a proposal to make SAC=1, SAM=00 to indicate the >> Unspecified Address. > >So why don't we stick with the way this was done in -04, not risking >any incompatibilities with ISA100? >(Of course, we should wait until we fully understand the ISA100 >situation.)
There's no incompatibility with ISA100.11a there. ISA100.11a only uses the stateful encoding format - or no compression at all. - ISA100.11a does not use the unspecified address. - It does not do any ND/DAD. - It always compresses the IPv6 address. In more details: - All of the traffic (but ne flow) uses the 16bit short addresses. A system manager assigns the short address and the mapping into an IPv6 address, a bit ala DHCP. Any remote address will be mapped into a local short address as well to be expanded by the backbone router. So ISA100 is fully stateful, and the short address is an index in the context - there's only one context. - The only exception is the join process. The joining of a new node is a special flow and the only one that uses a 64bits MAC-64 Link Layer Address. From the network layer standpoint, the flow is performed using the EUI-64 based Link Local Address and should theoretically be encoded in a stateless form. But for simplicity ISA decided to have an implicit state virtually indexed by the MAC-64 link layer address and thus uses the stateful encoding even there. The 16bit address and the corresponding IPv6 address of the node are assigned as part of the join process. - Any flow is started under the control of the system manager that will pass the addresses of the remote node and other contract information such as QoS parameters. QoS parameters end up encoded in the flow label. DSCP in not defined at ISA100.11a so TF bits can be either 01 or 11. There's work at ISA100.15 and ODVA that aim at defining industrial DSCPs. When that happens, we might very well see this work retrofitted in a revision of ISA100.11a that would enable all the TF possibilities. CC'ing Christos who was the lead editor of the network layer section. Christos, do I miss anything? Cheers, Pascal _______________________________________________ 6lowpan mailing list [email protected] https://www.ietf.org/mailman/listinfo/6lowpan
