On 19 Apr 2010, at 18:40 , Brian E Carpenter wrote: > Yes, but *my* rather confusing reply should really have said that > although details have certainly changed since 1994, we still don't > *know* whether a newly defined IPv4 option will or will not survive > a trip across the Internet on any particular path.
I don't *know* whether any of my packets will survive a trip across the Internet. Earlier today, the probability of a packet leaving my site (or entering it) was close to zero. Right now, the probability is vastly closer to 1, but it is certainly not equal to 1. ISP folks are telling me that the presence of IPv4 options is *ignored* by the IP backbone routers. Put in other words, IPv4 packets containing options *are* delivered across the backbone -- they aren't dropped by the backbone router and they aren't slow-path forwarded by the backbone router. So as near as one can tell, what has happened is that all IPv4 options are "end-to-end options" -- at least within the backbone. The real issue is probably security gateways. At least some security gateways have configurable options whether to look at (some) IPv4 options and to make some set of decisions about packet handling based on those options. [For an example, see RFC-1108 which is more widely deployed today than 20 years ago, by the way.] The good thing about security gateways is that they are deployed by end sites. This means the end site has control over the security gateway configuration, and that means that new protocols (e.g. SIP) and other new technologies (e.g. an IPv4 option) can be deployed if the site finds it desirable to do so. Cheers, Ran _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
