Re: [Int-area] The small address use case in FlexIP

[Stewart Bryant <[e-mail]>]

> On 5 Feb 2021, at 12:06, Jiayihao <[e-mail]> wrote:
> 
>  
> - The header format could be described either in separate draft or be included in previous draft. The reason we have not provide a header format yet is that the address format itself is already a complex topic, so it's better for us to discuss the address first (as well as the problems and gaps), thus we can have a better understanding if a flexible address structure is a promising way to go.

Again I disagree, you cannot entirely divorce the discussions.

The problem has to be solved as a whole and decisions about the packet effect decisions about the address structure. For example if the new packet is much larger than IPv6  the case for short addresses becomes even weaker.

If I look at the address design that you have proposed this is dominated by the short address constraints. If short addresses (in the physical sense rather than a logical sense where they are the suffix of a larger address) turn out to be a trivial efficiency saving, then we would almost certainly use another address design.

For example <len><family><address> as used in ISO 8474 is an extremely extensible format at a cost of a two byte overhead, and would allow the address designers to proceed independently from the packet designers and develop new address types during the deployment lifetime of the protocol. This is unlike the approach in the drafts we are discussing where a lot of discussions are proposed up front and where the development of the address architecture is significantly constrained.

Best regards

Stewart