Re: [rohc] ROHCv2 Packet Formats

[Klaus Warnke <klaus.warnke at acticom.de>]

Hi Anil,

At Wed, 11 Nov 2009 10:22:40 +0530,
Anil Maguluri wrote:

> How to send IPv6 Extension Headers (Destination Option, Routing
> header, Hop-by-Hop options) in ROHCv2?
>
> The above extension headers are classified in static and dynamic
> fields.  There is no irregular fields in those headers.

Yes.  Btw, RTP has also no irregular fields, that is, the
rtp_irregular format is empty.

> 1. Do we need to consider static fields while selecting the flow?
>
> For example: In case of Destination option, next_header and length
> are static fields. If the length value changes means is that is new
> flow?

Yes.  The only specified way to transmit the next_header and length
field is the dest_opt_static format. And, see 5225 6.5: "The static
chain is only used in the IR header format.".  Therefore the only way
is the IR packet. And the IR initializes the whole context. But you
can (should?) use the same CID, if you are sure, that is the same
flow. From decompressor point of view it makes no difference. I think,
this szenario occurs seldom, so there is no smaller packet defined to
transmit such changes. In RoHCv1 the extension headers were compressed
with the so called table based compression. I implemented that, and I
can say it was very, very difficult for me. On one hand, as in RoHCv1,
you can define a difficult to implement feature to save some octets
for a seldom ocurring szenario with a high risk, that the
implementations are not interoperable, or, as in RoHCv2, an easier to
understand and implement mechanism, which wastes some octets. That is
my point of view.

> 2. How to send (in which packet) if the dynamic field value changes?
>
> For example: In case of Destination option, value is dynamic. If the
> value changes, on which packet do we need to send value information?

The dynamic chain is used in the IR and co_repair packets: "The
dynamic chain is only used in the IR and co_repair header formats.".
Therefore you can use both of them. The co_repair format is intended,
if a context damaged is assumed, but I think you can use it also for
that szenario. I think, I prefer using the IR header.

I hope, that helps.

br
Klaus