Re: #385: HTTP2 Upgrade / Negotiation

Hi Patrick,

On 10/24/12 3:51 PM, Patrick McManus wrote:
> I'm not an expert here, but yes, there are lots of ways to badly
> design your DNS and incur delays - but there are also reasonable ones.
> This is not substantially different than the 'out of baliwick' DNS
> issues from the 90's right? but that also shows it is resolvable
> within the current record type framework, as you say in this case by
> using the same name which is what I think we're all envisioning.

You're right in that none of this is new.  The nefarious example I gave
wasn't under control of the administrator of, but rather
the bad guy.  That's why there's a fair amount of code in many
implementations to avoid accepting additional information in any
authoritative way, but that may well cause additional lookups, defeating
the purpose of the shortcut.  It might be possible to race lookups in
some heuristic or predictive manner, but it may be just as well to
address this with a new record.

My suggestion is to ask the DNS directorate for a view.  They could say
that I'm in the weeds, but then I doubt it.  I think they would say that
we should carefully consider our requirements and available engineering
choices (buy/build).  Yes, SRV is there.  No, it's not that hard to do
another record, but it does take time to get out there, should we go
that route.


Received on Wednesday, 24 October 2012 18:11:00 UTC