On Fri, 2006-07-04 at 18:01 +0200, Freek Dijkstra wrote: > jamal wrote: > > Excellent point! It was bothering me as well but i couldnt express my > > view eloquently as you did. > > Explicitly defining scope may indeed be a better approach at first > sight, but there are a few very practical reasons not to use them. > > First, it would not be useful in this purpose. IPv6 link-local addresses > (FE80::0/10 addresses) are "local" in scope, just like the IPv4 > link-local addresses, but ARP replies for IPv6 should not be broadcasted > (since there are no address conflicts due to the larger address space). >
I think you meant ndisc which is a different code base than what the patch provided. So yes, no conflicts in V6 - and therefore this is unneeded. But not so in ipv4. > Secondly, this is matter of choice: either the IP address or the "scope" > parameter is authoritative when it comes to specific behaviour of > network devices. It'd say that since the IETF and IANA specifically use > IP addresses, we should follow that guideline. > This does not at all preclude what the IETF or IANA sets. The difference is mechanism vs policy. Should we hard-code checks for RFC 1918 IPV4 addresses in the kernel? After all, they are well defined by the IETF. The answer is no. This follows the same logic: The controller for v4 link local addressing sits in user space. It sets the values in conformance to to the RFC. > > To elaborate on my last point: > The alternative would be to require all application that set an IP > address to also set the scope. Indeed. What applications are you referring to? As far as i know there's a handful of competing apps which are vying to be the "zero config" daemons on Linux. > This is currently not done. Its a 33.333 seconds fix. > Adding three machine instructions is just the easier approach. I am not objecting to "some fix" in the kernel. I am objecting to the patch that was sent. If one was to set the scope to be link local, wouyld it work just fine? A patch may still be needed with those same number of instructions. The only question is: will it check for an ip address value or an ip address attribute? From a mechanism perspective, attribute sounds a lot reasonable. Lets see if it is even needed if you set the right attribute. [Deleted the rest of the arguement since it resolves around the two above] cheers, jamal - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html