Hi Thomas, I would again suggest that instead of making it non-mandatory, we could provide a seperate set of requirements - for different device types.
OSPFv3 currently uses IPsec because the assumption is that IPv6 mandates IPsec, and that means we do not need any other mechanism for the same. OSPFv2 (normal OSPF) uses its own internal mechanism for authenticating protocol packets. The idea is that if you change such requirements, it may break the assumptions taken by the other protocols regarding the same. Thanks, Vishwas On Tue, Feb 26, 2008 at 8:18 AM, Thomas Narten <[EMAIL PROTECTED]> wrote: > IMO, we need to get over the idea that IPsec is mandatory in > IPv6. Really. Or that mandating IPsec is actually useful in practice. > > It is the case that mandating IPsec as part of IPv6 has contributed to > the hype about how great IPv6 is and how one will get better security > with IPv6. Unfortunately, that myth has also harmed the overall IPv6 > deployment effort, as people look more closely and come to understand > that deploying IPv6 doesn't automatically/easily yield improved > security. > > We all know the reality of security is very different and much more > complicated/nuanced then just saying "use IPsec". > > Consider: > > IPsec by itself (with no key management) is close to useless. The > average person cannot configure static keys, so the result is (in > effect) a useless mandate (as a broad mandate for ALL nodes). > > What applications actually make use of IPsec for security? A lot fewer > than one might think. For many IPv6 devices/nodes, if one actually > looks at the applications that will be used on them, they do not use > IPsec today for security. And, there are strong/compelling arguments > for why IPsec is not the best security solution for many applications. > Thus, requiring IPsec is pointless. > > To be truly useful, we (of course) need key management. If we want to > mandate key management, the stakes go way up. IKEv1/v2 is not a small > implementation effort. And, we are now in the funny situation where > IKEv1 has been implemented, but due to shortcomings, IKEv2 has already > been developed. IKEv2 has been out for over 2 years, but > implementations are not widespread yet. So, would we mandate IKEv1 > (which is obsoleted and has documented issues), or do we mandate > IKEv2, even though it is clear it is not widely available yet? > > IMO, we should drop the MUST language surrounding IPsec. The technical > justification for making it MUST are simply not compelling. It seems > to me that the MUST is there primarily for historical/marketing > reasons. > > Note that dropping the MUST will not mean people stop implementing > IPsec, where there is compelling benefit. Indeed, note that the USG > has already moved away from IKEv1 and has strongly signalled that it > will require IKEv2 going forward. So I am confident that IPsec (and > IKE) will get implemented going forward. > > But there is no reason why IPsec should be mandated in devices where > it is clear (based on the function/purpose of the device) that IPsec > will in fact not actually be used. > > As a general "node requirement", SHOULD is the right level, not MUST. > > Thomas > > > -------------------------------------------------------------------- > IETF IPv6 working group mailing list > ipv6@ietf.org > Administrative Requests: http://www.ietf.org/mailman/listinfo/ipv6 > -------------------------------------------------------------------- > -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: http://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
