[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [rohc] Liaison towards 3GPP
"Casati, Alessio (Alessio)" wrote:
>
> > I have also seen these not yet concretized proposals for header
> > stripping and I don't think that they are a preferred way.
> >
> Tdoc 2-00-771 from Nokia (sorry for introducing company related
> things here, which I hate to do, but this is how SMG2 works) is a good
> document, although not yet finalized and approved, for instance.
> Also There are some merits in handling RTP play out at the RNC
> for R'99 CS backward compatibility.
>
> > They violate
> > the end-to-end integrity of IP. To me it seems like these proposals aim
> > at "eating the cake and still try to keep it".
> >
> For the sake of truth, they don't violate IP integrity,
> but RTP integrity (not to mislead anybody, this is not WAP
> based ;).
Are you sure about this. If the IP ID is non-sequential as it commonly
is, I am pretty sure it will be stripped and reproduced without
preserving the original value. Hence, violation of IP integrity.
>
> > Furthermore, I seriously doubts that there is any large
> > gain in capacity having this kind of algorithms. The proposals I've seen
> > involves sending some header related information over the air-interface,
> > hence they will create some overhead. Since there are no actual detailed
> > proposals on header stripping (to my knowledge) it is not possible to
> > estimate the overhead, however, it is not unlike that it would be some
> > 0.x of a byte overhead in average per packet.
> >
> I would say that once the bearer is set up, maybe that X
> is very close to 0.
Radio bearer setup would include extra overhead for setting up the
stripping scheme, I think you should count these bytes as well.
>
> > Given the propsals for header compression in ROHC wg today, the
> > difference in efficiency will not be large. ROCCO for instance is
> > currently able to compress the header down to an average of appr. 1
> > byte.
> >
> 1 byte out of 20 is 5 %.
1 out of 30 something is probably more realistic, => 3 %.
>
> Spectrum license B in UK auction is now valued around $6 billions
> 5% is roughly valued $300 M even 30% of that is still $100 M. That's enough
> to start looking at the merits of header stripping.
3 % of bandwidth increase does not map linear to 3 % in terms of system
spectrum efficiency, especially not for TDMA systems such as EDGE. It
will
probably be a smaller decrease in spectrum efficiency.
>
> > I don't expect the final scheme to be any less efficient. Hence,
> > with header compression the efficiency will be only slightly smaller,
> > but you have the very desirable property of IP end to end.
> >
> when trading IP end to end for voice services and $300 M (or $100 M) a
> carrier may like the latter. This however does not mean we don't need
> the ROCH stuff. People just must be informed and be aware of the
> context they are working in (I assume not everybody here has visibility
> on everything, and I would like to be informed about what I can't see
> myself)
I fully agree.
/ Krister
__________________________________________________________
Krister Svanbro, M.Sc. - Ericsson Research
AWARE - Advanced Wireless Algorithm Research at Erisoft
Box 920 | Tel. : +46 920-20 20 77
SE-971 28 Luleå | Fax : +46 920-20 20 99
Sweden | GSM : +46 70-621 69 42
mailto:Krister.Svanbro@epl.ericsson.se