Re: RFC 4861 missing updated-by

[Brian E Carpenter <[e-mail]>]

On 17/08/2017 00:48, Mikael Abrahamsson wrote:
> On Wed, 16 Aug 2017, Timothy Winters wrote:
> 
>> I looked thru the IPv6 Ready Logo Test Specification for this, we test 
>> this for NAs and NSs but not RSs/RAs.  I've filed a bug to update the 
>> Test Specification to check this in the next release.
> 
> Great takeaway from this discussion. I have personally run into problems 
> with an implementation that did not ignore MBZ bits but instead required 
> them to be 0. Since the sender of the packet had a bug and didn't zero 
> them, this caused things not to work.
> 
> The sender of the packet which didn't zero the bits immediately understood 
> what the problem was and offered to correct things, however the vendor of 
> the device that didn't ignore the bits took a lot longer to convince that 
> they were doing something wrong.

Sorry to repeat myself, but that's exactly the point we made in
https://tools.ietf.org/html/rfc6709#section-4.2
In other words, it's a quite general problem, and an excellent area
for test suites to exercise.

And IMNSHO it clearly shows why we need an "updates" for such cases, as
well as an IANA registry. Implementers make mistakes, and we should use
every tool we have to make this less likely.

   Brian

> 
> This was OSPFv3 and it had a 24 bit field in there with the remaining 8 
> bits being MBZ and ignored. The vendor put this into a 32 bit field and 
> compared meaning that when the other end messed up their zeroing the 
> reserved bits, the compare failed. I'd imagine this is not too uncommon 
> problem.
>