On Fri, Sep 15, 2023 at 12:30 PM Salz, Rich <rsalz@xxxxxxxxxx> wrote:
> The key point here is that all that is required to make use of SRV for Web
> Service lookup is to have some defined convention for converting the
> information from the SRV record and accompanying TXT record into some form
> of URI.
And how is this different from adding a new RRtype? Seems to me you need to change everyone both ways. I mean yes, this could be done in a single resolver library, but last I looked there were lots of those.
There is an RR type that is essentially 'SRV and a little bit more' already - NAPTR. It never took off (outside SIP) because there is a first mover problem in deploying new RRs and because regular expressions are well, regular expressions.
A very, very large number of Internet users use a DNS service provided by their registrar which is accessed through an interface that is limited to A, AAAA, SRV, TXT, MX and a few others. It took over a decade to add CAA to that list.
My 1&1 account is of this type. Now I do have a full DNS on another hosting provider but that took me half a day to configure. So that is a very expensive additional step to demand of people before they implement some new application or service.
I really don't see how introducing a new RR is better than the SRV/TXT approach which is what the market has already decided on. The new RR vs SRV/TXT debate has been over for a decade, the only place it is not over is in IETF circles where the DNS cabal has its own agenda and flatly refuses to accept the fact that using a new RR is a problem.
So the choice is between
1) SRV/TXT, works now, is the defacto approach already widely adopted albeit with not quite as much consistency as one would like
2) Some new RR which will take many years to roll out and offers no advantage to the deployment community.
I know the DNS cabal really really thinks new RRs are the way to go. But the market clearly disagrees.
The other issue that isn't mentioned is that the only way to access DNS information besides an A /AAAA record is to write your own DNS library or use some third party code. So even if you want to use SRV, you have to be writing client side code.
If the IETF was to fully get behind the SRV/TXT approach then we could hope to see nslookup being supplemented/replaced by a platform level replacement that works the same everywhere in a way that third party afterthoughts never quite achieve.
