Fernando Gont <ferna...@gont.com.ar> writes: > On 21/01/2011 04:49 p.m., Thomas Narten wrote:
> > But other uses of the Flow Label outside of that do not have that > > requirement. > > > > So, if you do traditional flows, you have to be sure to generate Flow > > Labels carefully. The default usage outside of signaling does not > > have the same requirement. > Having two different FlowLabel generation algorithms does not sound like > a good approach to me (not that you're proposing that, though) -- in > particular, when we'd have to work on an API for apps to select the > proper one. You misunderstand me. For flows where the application wants special service, you can afford to have a more sophisticated algorithm. For the vast majority of applications, they won't set the flow label and will leave it to the OS to use heuristics (like each socket gets a different value). That is the case we are trying to make work with load balancers. These can use a much simpler default algorithm. And we need to allow for *very* simple devices, like sensor devices, etc. > IMHO, pick an algorithm that is suitable for all cases, and that's > it. I think it's fine to have a simple default. What I'm arguing against is making the default case more complicated than it really needs to be. We can have a more sophisticated algorithm in those cases where it actually matters, since those will not be the common case (if indeed, we ever develop standards in this area.) Thomas -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
