Re: [secdir] review of draft-ietf-isis-bfd-tlv-02

["Les Ginsberg (ginsberg)" <[e-mail]>]

Sam/Stephen -

> -----Original Message-----
> From: Sam Hartman [mailto:[e-mail]]
> Sent: Sunday, July 25, 2010 4:19 AM
> To: Stephen Kent
> Cc: [e-mail]; Les Ginsberg (ginsberg); [e-mail];
> Stewart Bryant (stbryant); Chris Hopps (chopps)
> Subject: Re: [secdir] review of draft-ietf-isis-bfd-tlv-02
> 
> >>>>> "Stephen" == Stephen Kent <[e-mail]> writes:
> 
>     Stephen> The Security Considerations section is just one
paragraph,
>     Stephen> which states that the addition of this feature does not
>     Stephen> adversely affect the security mechanism (sic) of IS-IS.
> I'm
>     Stephen> not questioning this assertion, based on reading this
>     Stephen> document, but I think a couple of additional sentences
are
>     Stephen> needed here, to justify the assertion.
> 
> 
> How about something like: This TLV does not adversely affect the
> security of IS-Is. The primary consequence of using the BFD mechanism
> in
> a case where it is not supported is incorrectly detecting a false
> failure of bidirectional forwarding.  The primary consequence of not
> using BFD when BFD is supported is failure to detect situations in
> which
> bidirectional forwarding is not happening on a link. At worst, this
> leads to a denial of service condition. However, a party who can
> manipulate the contents of this TLV is already in a position to create
> such a denial of service by disrupting IS-IS routing.

The mechanism introduced by this draft is used to determine when/if to
use BFD session state as a prerequisite to forming an IS-IS adjacency
and/or to trigger IS-IS adjacency state transitions. It does not make
any changes to the operation of BFD itself - which I think your wording
may unintentionally imply. How about:

"The TLV defined within this document describes an addition to the IS-IS
Hello protocol. Inappropriate use of this TLV could prevent an IS-IS
adjacency from forming or lead to failure to detect bidirectional
forwarding failures - each of which is a form of denial of service.
However, a party who can
manipulate the contents of this TLV is already in a position to create
such a denial of service by disrupting IS-IS routing in other ways."

??

> 
> If the above paragraph is correct, I think it may go some distance to
> addressing Stephen's concern.
> 
> However, I wonder how the authentication mechanisms of BFD interact
> with
> the authentication mechanisms of IS-IS? 

They do not interact.

> Is it possible to get into a
> situation where IS-IS is authenticated but BFD is not?  If so, that
> should be discussed in the security considerations section.

It is possible to use BFD in conjunction w IS-IS today - and of course
folks are actually doing that. The mechanism introduced here does not
change that - it just allows the protocol to operate correctly when
IS-IS PDUs and IP/IPv6 packets do not share fate. I don't see that we
have introduced any change in the way authentication is used/not used
for either IS-IS or BFD - nor have we introduced the use of BFD in
conjunction w IS-IS. So I don't see that discussing the point you raise
is needed.

   Les

> 
> --Sam