If I recall correctly, the policy table can list prefixes explicitly in priority order. So for mh, than means that policy can say: if A fails, try B; if B fails, try C.
But I may be misremembering 3484, it's been a while...
Brian
Hi Brian, Arifumi,
I am not sure that the policy table falls under this particular component, since as i see it, the policy table is more related with locator selection when there are no failures and it is used as a mean to express administrative prefferences.
Anyway, as i understand it, Arifumi is considering mechanisms to distribute RFC3484 policy table, so this is not really related to locator selection, but rather with centralized policy management.
IMHO this will be a requirement, especially as the sites grow bigger, because manually configuring the policy table of each hosts doesn't seems a reasonable approach when there is a considerable number of hosts.
Well, in any case, i can see basically two options to provide automatic configuration of RFC3484 policy table:
- router Advertisement
- dhcp option
the porblem with RA is that the full policy table has to fit in a single RA message, since imho the distribution of the policy table should be atomic (having only part of the policy table configure may cause undesired behaviour), in addition, RA requires the configuration of the policy table in at least one router per link, which may be cumbersome
the problem with dhcp is that you need a dhcp server, but it provides most of the desired features AFAICS
imho, a policy distribution mechanism will be needed at some moment.
We have made a couple of implementations in linux of the two policy table distribution mechanisms, one using dhcp and another one using RA, to see how this work.
If anyone is interested, we can discuss it.
Regards, marcelo
El 12/08/2004, a las 16:26, Brian E Carpenter escribió:
One of the components we have already identified (see the minutes from San Diego) is
Locator selection after a failure has been detected / Choose new address pair
RFC 3484 policy distribution might be a possible solution for that component. We aren't chartered to do basic design in this WG, but it would certainly be good to see proof-of-concept for this (and every) component.
Brian
Arifumi Matsumoto wrote:
Oops, I forgot to send CC to multi6. -- Hi marcelo,
Well, even if you do a per host locator selection you can still manage policy in a centralized fashion, if we define proper mechanisms to perform policy distribution, for instance a dhcp option for distributing RFC3484 policy table or something similar. However, i don't know how you can enforce policy in a centralized manner when locator selection is performed by the end hosts.
I believe that such a mechanism for policy distribution is really important. By making use of the existing RFC3484 framework, IMHO we can develop much manageable multi-home environment in a simple and easy manner. Now I'm working on this mechanism to publish a new I-D. But I'm not sure such kind of topic is appropriate for this WG or not. I believe this part of the technology should be necessary for some of the proposals discussed here, though. -- Arifumi Matsumoto Ubiquitous Computing Project NTT Information Sharing Platform Laboratories E-mail: arifumi@nttv6.net