"Perry E. Metzger" wrote:
>
> Brian E Carpenter <[EMAIL PROTECTED]> writes:
> > Let's either clarify them, as in the one para summary that Tony Hain
> > sent a few days ago, or simply write the field off as reserved, in which
> > case IPv6 will have no advantage of IPv4 for QOS purposes.
>
> I think the notion of making our protocols better for QoS purposes is
> an admirable one. My problem is that I don't think anyone has well
> demonstrated that the flow label *will* give us an advantage for QoS
> purposes. I understand the arguments. I just haven't seen anyone
> step up with the crucial "we at large router vendor have done the
> experiment and even though we have to reach into the packet anyway to
> do tuple extraction having this in a percentage of the packets will
> increase performance in this quantitative way."
Chicken and egg. People building protocol silicon aren't going to
commit to the flow label until it has a useful definition.
Also, the goal is not to increase total performance; it's to discriminate
between different traffic streams in a quantitative way. That's done
today in plenty of products; the flow label would simply add a tool.
>
> I'll repeat (broken-record like), we've already deployed the end point
> implementations in the field. We can't change those stacks at this
> point.
No, so it's just as well the API supports the flow label :-)
Brian
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
