Wait, let me attempt to rephrase this... if you are using MD5, SHA-1 or whatever w/ sequence numbers, then it has to be meticulous. That definitely does not come across in Section 6. Here is my attempt to clarify the paragraph:
"If using MD5, SHA-1 authentication with or without ISAAC and with sequence numbers to control packet loss, BFD MUST use an authentication type (bfd.AuthType) that is of type Meticulous Keyed MD5 Authentication, Meticulous Keyed SHA1 as defined in BFD <https://www.ietf.org/archive/id/draft-ietf-bfd-stability-20.html#RFC5880> [ RFC5880 <https://www.ietf.org/archive/id/draft-ietf-bfd-stability-20.html#RFC5880>] or other authentication types that provide for meticulously increasing sequence numbers can also be used such as Meticulous Keyed ISAAC for BFD Authentication <https://www.ietf.org/archive/id/draft-ietf-bfd-stability-20.html#I-D.ietf-bfd-secure-sequence-numbers> [I-D.ietf-bfd-secure-sequence-numbers <https://www.ietf.org/archive/id/draft-ietf-bfd-stability-20.html#I-D.ietf-bfd-secure-sequence-numbers> ]." I will freely admit that the above is one hell of a long sentence... Feel free to fix that. The mixing and matching of something called NULL authentication with authenticated mechanisms is ridiculously complicated, and not at all clear in the draft until one gets to that one and only paragraph in Section 6. I can't wait until someone decides to implement it (sorry, it is almost meeting week and I might be cranky) Again my comments are not blocking. Deb On Wed, Oct 29, 2025 at 4:26 PM Jeffrey Haas <[email protected]> wrote: > Deb, > > Glad to hear we're largely converged. A minor clarification here may help > - or might suggest the need for a minor tweak to verbiage: > > > On Oct 29, 2025, at 4:16 PM, Deb Cooley <[email protected]> wrote: > > > > [DC] I really have no idea what the point of the second para is in > Section 6 (the rest of the subsections are fine, and the first > sentence/para in the section is fine too). Why does this even need to be > mentioned? There is literally no mention of any of these techniques > anywhere else in the specification. It appears here, out of the blue. > Personally, I'd delete it. But these aren't blocking comments, so if you > all think it is clear, then I'm good. > > > > In RFC 5880, we support the following authentication types: > > 1 Simple Password > 2 Keyed MD5 > 3 Meticulous Keyed MD5 > 4 Keyed SHA1 > 5 Meticulous Keyed SHA1 > > The distinction is MD5 and SHA1 also are permitted to operate in > NON-meticulous mode. The distinction is that the same auth seq# can be > used more than once. The motivation for that behavior is we still provide > BFD messages saying "we're up!" but we're only requiring authentication to > be re-done when the sequence number changes. Otherwise, you effectively > can do a memcmp and be satisfied that nothing has changed state-wise or > security-wise. > > For BFD stability purposes, the requirement is to use the meticulous mode > in order to provide detection of lost packets. Non-meticulous mode can't > give us that. > > With that explanation, is everything clear or do you have rewording you'd > find helpful? > > -- Jeff > >
