On 31 Oct 2026, at 19:13, John R Levine <[email protected]> wrote:

>> As author of RFC 7901, I see no issues with embedding dns chains that can 
>> be dnssec validated. Anything without validation is just too
>> vulnerable to attacks.
>> But I also don’t believe the won time would actually go to the user - they 
>> would go to the advertiser auction times to get the highest bid for
>> the ad served on the web content.
> 
> Wow, that's impressively cynical.  But you're probably right so I agree it's 
> not worth it.
> 
> I'm also wondering even it if were worth it, how often a web page would have 
> an embedded answer that wasn't already in the browser's local cache.

I think it's also worth considering whether the answer delivered directly by 
the HTTP server could usefully be different, not just whether it's already 
cached.

In environments where the DNS response adapts to the perceived needs of the 
client it is quite possible that the HTTP server could have a more useful 
answer, one that is possible to compute with less guesswork. Such "stupid DNS 
tricks" (a term of art, not intended to be rude) rely on all kinds of 
unreliable metrics to make decisions. We don't do things this way at Cloudflare 
(we use anycast to steer traffic) but other significant CDNs do.

A small client-side script embedded in an HTML object could perform layer-7 
tests (e.g. along the lines that is already and frequently done in the name of 
"real user monitoring") that might guide the addresses used for subsequent 
fetches much more effectively than DNS query metadata guesswork. That DNS 
guesswork is also subject to constant change since it depends on heuristics 
that are in the control of many other people, whereas moving the decision point 
to the HTTP endpoint is simplier and more deterministic.

For people who care about web performance, this might well be a bigger reason 
to be interested in embedded answers than DNS query latency.

I have no comments about the mechanics or economics of web advertising (or of 
Paul's cynicism :-) But I'll observe that web performance is a real thing that 
a lot of people put energy into.


Joe
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to