RE: [rohc] Implementation parameters and signals

[Mikael Degermark <micke@CS.Arizona.EDU>]

Hi Juha, 

see below.

At 02:40 PM 10/20/00 +0300, you wrote:
...
>-jk- Bits as channel identifiers? The actual channels in WCDMA are separated
>by other identifiers so these 5 bits are not needed for that purpose. What I
>originally tried to propose is that ROHC would provide compressed header +
>associated CID (CID not part of the compressed header). By having such
>possibility the logical layered architecture could be maintained (in 3GPP).
>By having the CID separate as an output from ROHC we would be able to get
>rid of 1 octect overhead in ROHC (in case CID can be provided by link) and
>be able to feed in several flows to ROHC but get them out as separate. I try
>to illustrate it in the figure below.
>
>-------------------------------------------------
>PDCP                |   n flows into ROHC
>                    V
>          -------------------------
>          | ROHC                  |
>          |                       |
>          -------------------------
>            |     |      |      |       1. Flow 1 'logical CID' 0
>            v     v      v      v       2. Flow 2 'logical CID' 1
>            1.    2.     3.     4.      3. Flow 3 'logical CID' 2
>                                        4. Flow n 'logical CID' n-1
>
>            Put compressed headers (+payload)
>            to PDCP PDU and copy 'logical CID'
>            to PDCP header (those 5 bits)
>
>                      |
>                      | all flows travel over same radio bearer
>                      v
>---------------------------------------------------
>RLC

It is not necessary to map the channel concept used by ROHC one-to-one to the
channel concept used by WCDMA. What I proposed was essentially that ROHC
uses those extra 5 bits to further divide a WCDMA channel into ROHC channels. 

This means that
1) no changes to the ROHC protocol is needed.
2) the interface you outline above can be used, i.e., in addition to the packets, also
the five bits are passed between ROHC and lower layers.
3) there needs to be some logic in the ROHC module that deals with the channel
identifiers. The change is minimal, however, essentially one needs to add a 5-bit
channel id to each context in use. 

This is very similar to what you proposed, the principal difference is that the 5 bits
are viewed as channel identifiers by ROHC, and not as CIDs. This avoids 
changing the protocol, and gives the same savings in bandwidth since ROHC
can use small CIDs and place one stream using CID 0 on each of the 32 channels.
There is also some extra flexibility, since by adding a single byte per packet, another
15*32=480 flows can be accomodated without renegotiation.