Re: [rohc] [Fwd: I-DACTION:draft-svanbro-rohc-lower-layer-guidelines-00.txt]

[wnk@fub.it (Roberto Winkler)]

Krister and all other HC-ers,

It's true what the sentence in latest Krister's mail says

>RLC and PDCP are layer 2 protocols in 3GPP and they don't provide with
>this kind of error detection. The error detection is instead done at the
>physical layer - layer 1.

but, please, consider that this indication is provided for an entire
Transport Block, which may contain the entire (compressed or not
compressed) header but also bits from the payload of the PDCP PDU
(depending on the data rate on the radio interface, the PDCP header length
and L2/L1 configuration). In this case the CRC status indication provided
by L1 to L2 is of no help in determining whether the header is ok or not,
as the possible errored bits in a Transport Block may pertain to the header
as well as to the payload of the PDCP PDU. The consequence is that L2 has
no means to provide the required service to header
compression/decompression (unless data rates are constrained to values that
ensure a 'perfect' mapping between the boundaries of the (compressed)
header and those of one or more radio frames (e.g. 10 ms worth of
information).

To summarize, I don't think that current 3GPP specs support our needs and
I'm not sure that we can go on by simply saying we in ROHC focus on
development of an efficient scheme and how to use it in UMTS it's a matter
of 3GPP. While pushing 3GPP to improve their specs, we should also take the
current status into account and work towards good solutions that fit the
best possible to the state of the art.

I understand these issues are not in the scope of this IETF group, but I
bet that many of us have UMTS and other 3G systems when dealing with header
compression strategies so please have patience....

I wish I could attend this IETF meeting.....

BR

Roberto



>Roberto,
>
>The purpose of producing a draft in the ROHC wg with lower layer
>guidelines is to push these kind of discussion and work to the
>standardisation bodies were they belong. This allows us in ROHC to focus
>on development of an efficient scheme. E.g., the discussion of how to
>enable error detection in for example 3GPP type of links should thus be
>solved in related groups in 3GPP. It is however  important that we in
>the ROHC wg makes sure that the scheme is possible to use in relevant
>systems such as those standardized by 3GPP. However, I think that this
>specific question is an issue for 3GPP rather then the ROHC wg.
>
>To give an short answer:
>RLC and PDCP are layer 2 protocols in 3GPP and they don't provide with
>this kind of error detection. The error detection is instead done at the
>physical layer - layer 1. If you look at the attributes of an UMTS
>bearer (in 3GPP 23.107 for example) you'll se that there are parameters
>for residual bit error rate, etc. It is possible to choose a certain
>value for residual BER, which results in that a CRC of for example 8 or
>16 bits is used for error detection. Moreover, in release 99 of 3GPP is
>already now RFC2507 supported and 2507 requires also error detection on
>headers.
>
>Hence, I am confident that 3GPP type system can provide with the
>required error detection but we must put requirements on this error
>detection.
>
>/ Krister
>

Roberto Winkler
Senior Researcher
Fondazione Ugo Bordoni
Via B. Castiglione, 59
00142 Rome Italy

tel +39 0654803411
fax. +39 0654804404
e-mail wnk@fub.it