On Thu, May 6, 2021 at 4:53 AM Sara Dickinson <[email protected]> wrote:

> > Section 7:
> >
> > * "Implementations SHOULD use [RFC7828] [...] to manage persistent
> > connections." -- since this is a SHOULD, can you suggest why an
> implementer
> > might opt not to do this?  Section 7.3.1 does a great job of handling the
> > SHOULD it presents, for example.
>
> Implementations typically include an option to use fixed timeouts on
> either side of a connection for simplicity which can work just fine for
> e.g., low transfer rates - I could add text to that effect?
>

Yes, something that rounds out this SHOULD story would be great.

> Sections 7.3.2 and 7.3.3:
> >
> > * Same issue as above with the SHOULDs here.
>
> Short answer, backwards compatibility.
>
> For 7.3.2 there was a discussion in the working group about the
> significant code complexity involved for several implementations to meet
> all 3 requirements (particularly the second two points).  Since the
> performance improvements gained from these features is desirable, but not a
> critical for XFR, the discussion resulted in the SHOULD. Again, could add
> text on that if needed?
>

Adding something like that would be helpful, yes.

For 7.3.3, there is no previous specification for this behaviour and
> nameservers may behave differently today, hence the SHOULD here.
>

Same.

> Section 7.4:
> >
> > * "Therefore, it is RECOMMENDED that [...] there SHOULD be no more than
> [...]."
> > -- I don't know how to reconcile this RECOMMENDED-SHOULD combination in
> the
> > same sentence.  I suggest that the first clause can be dropped with no
> loss of
> > meaning.
>
> Double checking this I remembered that for consistency the text is
> actually lifted and tweaked from RFC7766 which it is updating which said:
> “It is RECOMMENDED that for any given
>    client/server interaction there SHOULD be no more than one connection
>    for regular queries, one for zone transfers, and one for each
>    protocol that is being used on top of TCP (for example, if the
>    resolver was using TLS). "
>
> So we could update it if you feel very strongly?
>

Alas, no.  If you're copying this from something that already has
consensus, it's probably best to leave it.  Thanks for checking.

-MSK
_______________________________________________
dns-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dns-privacy

Reply via email to