Re: [Http-srv] Alternative to SRV?

[Ray Bellis <[e-mail]>]

On 29/08/2018 19:10, Erik Nygren wrote:

> They unfortunately may be needed as otherwise we'd be back in the same 
> boat around CNAMEs at the zone apex.  At least for the CDN use-case, 
> once a non-trivial amount of information shows up in the rrtype, it 
> likely needs to be managed by the CDN.  Having "_443._https.example.com 
> <http://https.example.com>" means it can be a CNAME'd whereas if was a 
> record on "example.com <http://example.com>" we'd be back in the same boat.
> 
> The underscore labels also provide the ability to have multiple services 
> on the same hostname.

So does having multiple RR types.

> I haven't fully thought it through, but perhaps there is a better way to 
> support underscore labels and wildcards in a way that only authorities 
> need to worry about?  For example, allowing for "_443._https.*" records 
> in an "example.com <http://example.com>" zone in a way that authorities 
> could know how to handle?

IMHO that's exceedingly unlikely.   The wildcard label has special 
meaning and recursive servers also need to know about it for DNSSEC.

I'd like to hear from actual web content providers as to whether lack of 
support for wildcards would be an issue.

Having worked in that business (albeit many years ago) I suspect it 
would be.  There are many web services that offer content under 
<username>.<domain> where the username portion is synthesized from a DNS 
wildcard record rather than provisioning separate DNS records for every 
single registered username.

Ray