Re: [Bier] comments on draft-xu-bier-encapsulation-03

[Xuxiaohu <[e-mail]>]

Hi Tony,



Thanks a lot for your good suggestions. Please see my reply inline.



________________________________
发件人: BIER [[e-mail]] 代表 Antoni Przygienda [[e-mail]]
发送时间: 2015年10月31日 19:49
收件人: [e-mail]
主题: [Bier] comments on draft-xu-bier-encapsulation-03

Re-read and rethought. The key lookup will be pretty wide but if one can afford the TCAMs and would want to build something like this I guess it’s feasible.

Architecturally the information carried in this encapsulation is more than necessary (which is NOT good). The architecture clearly indicates that a sub-domain MUST match onto one MT-ID and with that carrying both is superfluous, carrying subdomain is good enough. This also resolves the issue of e.g. driving this encapsulation with idr-extensions which doesn’t even carry a multi-topology concept.

[Xiaohu] OK,  we will remove the MT-ID field from the BIER header so as to keep it consistent with the current BIER architecture.

As well, why does the encaps carry BFR-id ? is that source/destination/originator ? None of those seem necessary to me ?

[Xiaohu] BFR-id indicates the ingress BFR. It may be neccessary in some use cases such as performance measurement.

I think the suggestion towards the authors has been already extended to read the excellent  dataplane  design draft put out by the according design team recently. I support that. Occam’s razor in data plane goes a long way ;-)

[Xiaohu] Thanks again for your valuable suggestions.
Best regards,
Xiaohu


thanks

--- tony

“Your work is to discover your work and then with all your heart to give yourself to it.” --- Buddha