Re: ROHC and ULE

[Carsten Bormann <cabo@tzi.org>]

Hi Cedric,

a couple of comments.

I have made a couple of proposals at previous IPDVB meetings how to  
integrate ROHC into ULE.
E.g., see http://www3.ietf.org/proceedings/05mar/slides/ipdvb-0.pdf
The most important input that would be useful at this point is how  
ROHC will be used in actual systems, see below.


> Now, how having a multicast addressing to work  with ROHC remains  
> an open
> issue, but at least a basic solution would be the use of  
> unidirectional
> mode (no need to acknowdege)

ROHC works fine in U-mode on unidirectional channels.
On a satellite link, the high delay for a context repair may actually  
make this the best mode to look at.

> Negociation of parameters between compressor and decompressor
> This point does not appear to be address in the document yet, but  
> how to
> make the negociation is a real issue for the implementation of ROHC  
> over
> ULE. Probably some propositions should be discussed here ? (e.g.  
> use of
> default parameters according to the type of DVB network ? PPP  
> replacement ?
> use of SI tables ? etc...)

We could probably come up with some default values.  You would still  
need to look at capacity points such as number of CIDs.
These could be "announced" (as opposed to "negotiated").

> ROHC recommendation
> We strongly believe that one of the output of the task shall be  
> ROHC usage
> recommendations. This can be either parameters values proposition or
> analysis of the best suited mode  depending on different scenarios  
> (namely
> unidr system, satcom systems with a return link, in both mesh and star
> topologies)

I'm happy to do this work if enough people are interested.

> ROHC adaptation
> A interesting point to investigate can be ROHC adaptation to fit the
> satellite systems and the ULE stack. Two approaches can be  
> foreseen. The
> first one is dedicated to simplification of the compression scheme,

Actually, we have just done that:

draft-ietf-rohc-rfc3095bis-rohcv2-profiles-00.txt

"...The RoHCv2 specification introduce a number of
    simplifications to the rules and algorithms that govern the  
behavior of the
    compression endpoints..."

> and the
> second one to the optimisation of ROHC in this context.

see above/below...

> ULE over ROHC ?
> This has never been proposed, but maybe the compression gain could be
> pushed a little more, if new ROHC profiles integrating ULE (i.e.. x/ 
> IP/ULE)
> were supported ? What would be the main barriers for such an  
> approach ?

These would probably be easy to add to ROHCv2.

Gruesse, Carsten