Re: [Http-srv] Alternative to SRV?

> On 22 Aug 2018, at 5:40 am, Tony Finch <dot@dotat.at> wrote:
> 
> Ray Bellis <ray@bellis.me.uk> wrote:
>> 
>> As mentioned at the side-meeting in Montreal, I strongly believe that the way
>> forward should be a new RR that is specific for the use of HTTP(s) (c.f. MX
>> for SMTP) and that would be automatically looked up by recursive resolvers and
>> returned in answers [*]
> 
> What will the backwards compatibility story be?

You look for all 3 record types simultaneously.  If you get back NODATA for the
HTTP record you use the A and AAAA responses.  If you get back data for the HTTP
record and ignore the A and AAAA responses.

If the HTTP lookup fails (NODATA is not a failure) you MUST NOT use the A and AAAA
responses.

Site operators decide when to stop publishing A and AAAA record for old clients
where they are using HTTP like they no longer publish A and AAAA records for SMTP
but instead rely solely on the MX record at the zone apex.  For MX this took 10
to 15 years as it took a long time to replace MTAs.  Browser vendors have shown
that they can replace the client side in less that 3 years.

> In the early days, what advantages will there be for DNS operators that
> will make them want to use this record?

If the browser vendors can push out updates quickly enough it will just be a working
record.  The advantage will be to site owners in that they no longer have to move
their DNS to their CDN provider, they can just add the HTTP record to point at the
service name their CDN provider gives them.

> Tony.
> -- 
> f.anthony.n.finch  <dot@dotat.at>  http://dotat.at/
> Bailey: Northwest 5 or 6, backing west 5 to 7. Moderate or rough. Showers.
> Good.
> 
> -- 
> Http-srv mailing list
> Http-srv@ietf.org
> https://www.ietf.org/mailman/listinfo/http-srv

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742              INTERNET: marka@isc.org