RE: [rohc] New version of chapter 5.9

[zhigang.liu@nokia.com]

> Mine too.
> 
> However, given that the UDP checksum is rather large, I would 
> consider compressing it if we could come up with a way to 
> *guarantee* that its semantics is preserved.
> 
> I don't know of such a way at this point in time. The fact 
> that the UDP checksum covers the entire UDP payload makes 
> it hard to compress and also less than useful
> for applications which are resilient to residual errors.
> 
> Micke

I bet you're thinking about UDP Lite. I've thought about why 
the IP header checksum and UDP checksum are treated differently 
in RFC 2508. I came to the conclusion that we really have two
types of checksum: one for headers, one for the application data. 
I would prefer a design that clearly separates them, i.e. only 
application takes care of the checksum for the data it 
generates and the header checksum ensure the data is delivered
to the right application (if it is ever delivered). In the case
of UDP checksum, it really should cover only UDP header fields 
(plus some IP fields, which looks kind of optimization to me) 
and let application puts the data checksum (if it needs one) into
the payload. Although not a clear separation, UDP Lite is an
evolution approach I believe we should take.

Br,
Zhigang