Comments on Nalini et al's IPv6 EHs presentation

[nalini.elkins at insidethestack.com]

Fernando,

> For instance, compare that vs. all v6-enabled destinations from Alexa's top 1-Million sites.

For what it is worth, you do know that many of these end up going to the same handful of CDNs, right? 

That is why we intentionally did NOT do this.

We will be back next time with more results.   

Thanks,

Nalini Elkins
CEO and Founder
Inside Products, Inc.
www.insidethestack.com
(831) 659-8360






On Tuesday, July 26, 2022 at 05:53:16 AM PDT, Fernando Gont <fgont at si6networks.com> wrote: 





Hi, Nalini,

On 26/7/22 09:11, nalini.elkins at insidethestack.com wrote:
> Fernando,
> 
>>For the numbers to be meaningful to assess whether IPv6 EHs are usable
>>on the public Internet, you definitely need to measure against a large
>>number of destinations (and also vantage points).
> 
> Our original test was from:
> 
> Warsaw
> Toronto
> Melbourne
> Mumbai
> Frankfurt
> Seattle
> 

That's certainly not statistically significative.



> So, multiple continents, multiple transit providers.   How many cities / 
> continents would you like?

As many as possible? ;-)

For instance, compare that vs. all v6-enabled destinations from Alexa's 
top 1-Million sites.



> We can certainly talk to the RIPE Atlas people.

Jen had done that at the time, with similar results to those in RFC7872, 
IIRC.



> When you redo your testing, I suggest you also test to see where exactly 
> the packet is dropped.

That's in RFC7872, already :-)



> For example, is it dropped at the source, in a 
> transit network or a destination network.   Merely redoing testing 
> against sites which we know are likely blocking extension headers is 
> unlikely to produce different results.

Not sure what you mean.

The goal of measurements is to measure with as many destinations and 
vantage points as possible, since *that* is what can shed light on the 
usability of EHs on the public Internet.

The first important result is whether the packets get to the intended 
destination or not. If they don't, whether they are dropped by the 
destination AS or a transit AS. And if dropped at the destination AS, 
it's also interesting whether they are dropped by the final box, or 
elsewhere.

Note: Geoff also did measurements in this area, with similar results to 
those in RFC7872.


Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: fgont at si6networks.com
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492