Re: [ANCP] Sub Capability Negotiation

[Francois Le Faucheur IMAP <[e-mail]>]

Hi Aniruddha,

On 28 Apr 2009, at 06:54, Aniruddha A (anira) wrote:

>
> With sub-capabilities, I would like to propose that we use 1 bit for
> each value, and change the
> Length to 4 bytes to simplify encode/decode.
>
> i.e, From draft-lefaucheur-ancp-mc-extensions-01, Section 5:
>
> 	Capability Data (1 byte): The following values are defined:
>         +  0x00: Reserved
>         +  0x01: "Transactional Multicast"
>         +  0x02: "Transactional Multicast" and "Multicast Admission
>            Control without Bandwidth Delegation"
>         +  0x03: "Transactional Multicast", "Multicast Admission
>            Control without Bandwidth Delegation" and "Multicast
>            Admission Control with Bandwidth Delegation"
>         +  other values: Reserved
>
> Instead, use:
>
>         +  0x00: Reserved
>         +  0x01: "Transactional Multicast"
>         +  0x02: "Multicast Admission Control without Bandwidth
> Delegation"
>         +  0x04: "Multicast Admission Control with Bandwidth
> Delegation"
>
> If the AN has all 3 Multicast sub-capabilities, then the sub- 
> capability
> value will be 0x07

This scheme would seem natural if the capabilities were independent.  
However, here the capabilities are natural supersets of each other. eg  
a device that supports "Multicast Admission Control with Bandwidth  
Delegation" ALWAYS also supports "Multicast Admission Control without  
bandwidth delegation". A device that supports "Multicast Admission  
Control without bandwidth delegation" ALWAYS also support  
"Transactional Multicast".
Going with a 1bit-per-subcapability scheme would mean we'd have to  
"deal" with all the invalid/undesired combinations eg 0x02 should  
actually be illegal, 0x04 and 0x06 also.
The only legal combinations would be 0x01, 0x03 and 0x07.

This seems to introduce unnecessary cases to deal with. No?
Do you see that it buys us anything?

Thanks
Francois


>
>
>
> --
> Regards,
> Aniruddha. A