[v6ops] FW: Can you please review and comment on draft-xiao-v6ops-nd-deployment-guidelines

[Xipengxiao <[e-mail]>]

Hi folks,

With Brian's permission, I publish his review and our discussion here.  Your comments are welcome.  Thanks.  XiPeng 

> -----Original Message-----
> From: Brian E Carpenter [mailto:[e-mail]]
> Sent: Wednesday, October 20, 2021 12:48 AM
> To: Xipengxiao <[e-mail]>
> Cc: Sheng Jiang <[e-mail]>
> Subject: Re: Can you please review and comment on draft-xiao-v6ops-nd- 
> deployment-guidelines
> 
> Hi XiPeng,
> 
> I do not really have the expertise for a deep review, but the work 
> looks very complete.
> 
> However, I do have some concerns about moving away from a traditional 
> deployment, despite the known problems. Obviously, very big WiFi 
> networks are a bad idea, but in my experience small ones (and small 
> Ethernet segments) work well. In particular there are two issues:
> 
> 1. Multicast is needed in many case, for mDNS in particular 
> (especially if DNS- SD becomes a success). So isolation is a problem 
> unless that requirement is satisfied. A correct implementation of MLD works well, from my experience.
> (There are also incorrect implementations, unfortunately.)

[XX] I agree that multicast is needed in some places.  But in those places network admin's can choose not to apply host isolation and use traditional ND.  Section 3.2 step 4 clearly states this.

There are scenarios where multicast is undesirable.  In those cases, network admin's can apply host isolation.  Because this draft does not dictate a single solution, and allow people to choose different solutions based on different scenarios, I think it's useful.

> 2. I expect a lot of objections to UPPH. It really is a ridiculous 
> waste of address space, unless we can reduce the subnet size with a 
> prefix greater than /64, and that seems very difficult in the IETF.

[XX] While I acknowledge that giving each host a /64 may seem wasteful, in reality it may not be so bad, for 2 reasons (1) I heard that RIR will give /29 to an applicant without requiring special justification. This will provide 2**35=32 billion /64's - more than enough I would think (2) We are giving a /64 to each mobile phone.  If we can afford that in this scenario (with by far the largest number of hosts), we should be able to afford that in other scenarios (with smaller number of hosts).  Do you agree?

Thank you very much again for your review.  

XiPeng