Archive

Hi.

8484 servers specify their set of doh uris via uri templates.. which
certainly could involve params not in 8484. Such a service would require
those params.

As for whether a server should be liberal in receiving a uri that does not
match it's configured template ... That's a specific instance of a question
perhaps more hotly debated than doh itself..

https://tools.ietf.org/html/draft-iab-protocol-maintenance-03

8484 does not, and really cannot, proscribe the error handling for whether
the rest of the http server fixes the issue

On Sat, Jun 1, 2019, 8:36 AM Christoph <cm@appliedprivacy.net>; wrote:

> Hello,
>
> we run a public DoH service and while running and comparing
> different DoH server implementations in practice we noticed that
> some have a higher error rate than others.
> The error rate was measured by looking at the fraction of each HTTP
> response code.
>
> It turned out that those server implementations with a higher error rate
> just do stricter HTTP parameter checking. Not necessarily the content of
> the parameter but the list of parameters contained in the request.
>
> The result is that some DoH clients are not compatible with some DoH
> servers.
>
> The error condition is caused by DoH clients that have additional HTTP
> parameters in the request that is not specified in section 4.1 of
> RFC8484 in combination with DoH server implementations that enforce the
> absence of any additional parameters.
>
> To increase interoperability we reached out to (some) affected DoH
> client and server vendors but we would also like to hear opinions from
> this WG.
>
> Should DoH servers reject requests with a '400 Bad Request'
> response when the request contains HTTP parameters not defined in RFC8484
> or
> should DoH servers be less strict and simply ignore additional HTTP
> parameters and proceed with processing the relevant HTTP parameters
> required to provide an answer?
>
>
> thanks,
> Christoph
>
> _______________________________________________
> Doh mailing list
> Doh@ietf.org
> https://www.ietf.org/mailman/listinfo/doh
>
>

Re: [Doh] DoH client-server interoperability vs. strict HTTP parameter checking