[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: failure detection




El 29/08/2005, a las 8:24, Erik Nordmark escribió:

marcelo bagnulo braun wrote:

I agree that using the "preferred lifetime" for signaling that a prefix is no longer working would imply to change the meaning of the field, but wouldn't it be possible to use the "valid lifetime" (set to zero) to inform the hosts that a prefix is no longer available?

If you have non-shim6 hosts, then a valid lifetime set to zero might mean that they'll immediately terminate any tcp connections that use that prefix. I don't think we want to do that, because packets might start flowing again before e.g., TCP gives up.


So I think that if we are going to add an indication that a prefix isn't working, so that shim6 hosts can change the locator they use more quickly,

agree, but i was thinking also about non-shim hosts (sorry for not stating) this before) (i.e. unmodified hosts) and whether it would be possible to enable routers to detect that the link with a given ISP is down and inform hosts within the multihomed host not to use it anymore (especially in order to avoid trying with prefixes that we know are down for establishing new communications and reducing latency at establishment)


(in this case, because we are dealing with unmodified hosts the option would be to use an existent tool or not to provide the functionality, i guess)



we should add a new indication and not overload an existing one to have these semantics.

   Erik