On 11/23/2011 07:41 AM, Fred Baker wrote:
I think people should go off and implement *something* in a dual stack home routing environment and report back on what happens.On Nov 22, 2011, at 2:36 PM, Howard, Lee wrote:Ray asked for people to post drafts for anything other than OSPF, because without an alternative, it will appear that we have consensus on OSPF. I haven't posted a draft on RIPng, because it would just work the way it's designed.
Makes sense to me.
I have mentioned my fondness for babel earlier in this thread. It's 'RIP on speed', and unlike rip doesn't have the counting to infinity problem... and does something unique and useful in addition to that, supporting mesh routing, which I think is rather important for wireless-mostly networks.
Regrettably I do not have time at present to submit a draft on this topic, as I'm too busy actually making code work in cerowrt, fixing bufferbloat, etc. (I'm strongly encouraged, btw, things like the newfangled "byte queue limits" patch and the QFQ queuing discipline may actually solve a great deal of the problems we are having in this space)
The choice of routing protocol is the least of my problems!While I'm off on other topics, a version of NAT for ipv6 has been submitted to the linux networking mailing list...
Prefix delegation is of interest...I have wide-dhcpv6 doing pd for example, and I note that the latest release candidate of dibbler looks pretty good as active development resumed about a year ago. As for how well that can integrate with anything else, is kind of unknown.
Does anyone actually have a home router running ospf over ipv6?a couple more comments below. I'd like to make clear that I'm neutral - merely that ospf didn't do what I wanted when last I tried it.
olsr and batman do - except that the former requires two daemons at present to run and is weird on ipv6, and batman doesn't do ipv6, which reduces my own choices to... no, I'm not going to say it...
A few people said http://tools.ietf.org/html/draft-howard-up-pio-00 is no better than RIP, and we already have RIP in home gateways. Can any gateway vendor confirm whether RIPng is already in gateways?
RIP is not available for v6 in any home gateway I'm aware of. I could be out of date on that.
it's a one liner for babel to pick up a default gateway injected by dhcp or dhcpv6.Proposed alternatives are: * OSPFv3. It's heavyweight for home routing. We still need a way to find the border and inject default. It could be used for DHCP-PD.
* zOSPF. It requires resurrecting work. I don't how much work it needs, or how big the protocol is.The Internet Draft could be summarized as "Use OSPF, and<do this> for subnet allocation". If you don't like OSPF, you don't like zOSPF.* UP PIO. It's new work, and requires work. It's lightweight, and solves the border problem, but not addressing. * RIPng. It's fairly lightweight, and it exists. It solves only the routing problem.To my way of thinking, as a default protocol for very small networks, you could do worse. I personally prefer OSPF, but I'm a snob :-)
And me, AHCP + babel...
* MANEMO. It requires resurrecting work, is pretty lightweight, and solves addressing and border problems. If you argue that we should reuse existing protocols (per the architecture draft), your choices are OSPFv3 or RIPng. I really don't like OSPFv3 in the home--it's too much protocol, though if someone can tell me about memory footprint, that would be helpful.
An ipv6 capable home router needs a minimum of 32MB of ram and 16 MB of flash, at present. Inside of that are many other daemons that eat up over half that ram, and the rest is used for buffering.
It will be nearly impossible to find one that has less than 64MB in the fairly near future.
Home routers do not swap. However, when under memory pressure they can generally drop read only pages with high effectiveness...
You can dismember any of hundreds of home routers and see the various components in action. there are daemons for dhcp, dhcp serving, ntp, dns, ra announcements, and a web server, at minimum.
Any set of "inside the home" internal routing tables is likely to be so much smaller than the other required features in the device as to be not worth thinking about much.
Unless people are seriously considering running BGP on the home router...... I can't think of any other modern protocol that will eat up more memory than (for example) the web server, or wireless management daemon, at least, not until we get around to routing thousands of in-home movement sensors, or something like that.
This assumption would require testing, of course. Particularly the sensors.
If you're comparing to RIPng, "a lot more". It's a more complex program, and it not only stores a route table, it stores a link state database. I'd have to go look at something to say this definitively, but I've heard the phrase "order of magnitude" in discussions like these.I also prefer draft-baker-homenet-prefix-assignment, so we don't need OSPF for addressing. Any discussion? Lee This E-mail and any of its attachments may contain Time Warner Cable proprietary information, which is privileged, confidential, or subject to copyright belonging to Time Warner Cable. This E-mail is intended solely for the use of the individual or entity to which it is addressed. If you are not the intended recipient of this E-mail, you are hereby notified that any dissemination, distribution, copying, or action taken in relation to the contents of and attachments to this E-mail is strictly prohibited and may be unlawful. If you have received this E-mail in error, please notify the sender immediately and permanently delete the original and any copy of this E-mail and any printout. _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet_______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
-- Dave Täht
<<attachment: dave_taht.vcf>>
_______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet