Michael Thomas <[EMAIL PROTECTED]> writes:
>  > Yes, I have substantial doubts on this. First of all, people already
>  > have silicon that does this astonishingly fast. 
> 
>    Define "astonishingly";

Ran Atkinson says Extreme's implementation (among other things)
extracts the N-tuple in one clock cycle once the packet is in the
ASIC. I may be mis-stating his claim but you can feel free to ask him
about it -- it isn't like he's a shy guy.

>  > Fourth, I have yet to see any real
>  > explanation of what this is going to do for people in
>  > practice. Repeating my earlier request:
> 
>    Then you're hearing what you want to hear.

Nope, I'm not. I've gone through the whole discussion. No one has yet
stated precisely how they will use the field and what sort of speedup
it will mean for them in practice.

I want to see that clearly and distinctly explained well before we go
through the exercise.

I'm looking for statements from several router vendors that look much
like this:

"Big Router Company believes that it is important that we have the
flow label because by implementing it, our studies show we will be
able to reduce time on operation X to 175 microseconds. We intend to
produce the hardware that will do this. Operation X is critical for
packet forwarding because..."

However I haven't seen anything like that. All I see is lots of vague
posturing and comments to the effect that "well, we defined the field,
didn't we? why don't we use it?"

I see no one giving actual figures from experiments for the benefits
of the field. I see no solid engineering explanation of the use of the
field. I see no real claims from vendors that they badly want the
field and will put support for the field into ASICs any decade soon.

>  > and I look at
>  > the fact that there are a whole lot of v6 enabled systems already out
>  > there in the field that aren't going away, I feel like I have to start
>  > asking hard questions.
> 
>    "Whole lot"? Please.

Yes. Every single new PC shipping with Windows has v6 code in
it. EVERY SINGLE ONE. Amazing, isn't it.

The last major company that isn't shipping v6 support is Apple, and
that will be fixed in OS X very soon.

We defined a protocol and the vendors are now actually shipping it. It
is actually going out to the field. Unfortunately, that makes it a bit
late to make serious changes.

>    And adding a flow label to the kernel is *hardly* rocket science.

Pretty fucking hard to stop the rocket once the SRBs are lit.


--
Perry E. Metzger                [EMAIL PROTECTED]
--
NetBSD Development, Support & CDs. http://www.wasabisystems.com/
--------------------------------------------------------------------
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]
--------------------------------------------------------------------

Reply via email to