Re: Header Stripping

["Lars-Erik Jonsson" <[e-mail]>]

Hello !

[To: PILC and AVT, AVT included since original message appered at the
PILC-list]

Ali, the answers to your questions should be ABSOLUTELY NOT and PROBABLY
NOT.

Header stripping is in general something really ugly and such ideas should
not be accepted. However, there is a reason why ideas like that come up.
The reason is that in cellular environments, the header compression scheme
for real-time (e.g. CRTP) of today is not working very well due to the
problematic characteristics of the links. This was studied in the internet
draft "draft-degermark-crtp-cellular". New header compression solutions are
needed that are efficient and robust enough against packet losses,
thereby making things like header stripping uninteresting.

A new concept for header compression, ROCCO, has been proposed and the work
can be studied at:
www.ludd.luth.se/users/larsman/rocco

It is important to speed up the progress of the header compression
development and standardization before ideas like those about header
stripping has come to far. Support for and participation in the ROCCO
development is needed

/Lars-Erik Jonsson


>Question:
>1. Should a recommendation for header stripping/regeneration be allowed
>in any IP network?
>2. Also, should this issue be dealt with in the performance
>implications draft?
>
>Background:
>
>I was just reading a Cellular network proposal (the name of group or doc.
is
>not important) for an IP network. For air-link optimization,
>header-compression is recommend. Fine. In addition an option for
>header-striping and regeneration down-stream is also allowed.  A statement
>in the doc. states,
>
> "Reducing of the header size is done by removing redundancy in the
>originally coded header information AND/OR REMOVING HEADER FIELD
INFORMATION
>AND THERBY LOSING FUNCTIONALITY"
>
>This disturbed me. I want to make a contribution to the group to delete
such
>text, but would first like  to get some feedback from this mailing list.
>
>Note: the doc. does state that these options are to be investigated and a
>final recommendation made. So header-stripping option is not set in stone,
>yet.
>
>Irfan Ali
>Motorola