DANE uses TLSA records like _443._tcp.skuldwyrm.no and _443._tcp.www.skuldwyrm.no which could also be used to indicate that https is available and thus the browser should try HTTPS first.

HSTS is the approach undertaken by browsers today, but this is a topic of active discussion.

If a SRV DNS entry is found indicating https then the browser can act as if the domain was in the HSTS preload list.

I guess this would need to be standardized into HTTP/2 though.

Speaking as a web developer it's super easy and quick to add this, alternatively some kind of TXT DNS entry could be used I guess. Google does this already for it's webmaster tools to authenticate site owners. The adoption could be rather high.

Here is a proof of concept (it's really quick'n'dirty'crude and I don't really know DOS batch so apologies for bad coding):

Attached is concept.bat which is a normal windows batch script.

Run the script as follows:
concept.bat skuldwyrm.no
concept.bat example.com

skuldwyrm.no should be opened in the default system browser using https protocol while example.com should open using http protocol.

Ideally the script should check the returned port and srv hostname and use those instead and a proper browser implementation I'll assume would too.

This lookup only need to be done the first time the browser encounters the domain and should cache it.

Deleted: concept.bat 524 bytes

concept.bat 524 bytes View Download

Looking up info on DNSSEC it seems that SRV is not authenticated, this is a shame.
TXT records are however supported by DNSSEC in which case I'd like to suggest the use of: hsts=domain

I have added a TXT record to my site as an example
TXT hsts=skuldwyrm.no

Note that concept.bat attached above uses SRV and not TXT, if a proof of concept for TXT is desired let me know and I'll see if I can do it in DOS Bash or Powershell.

Testing using more tools shows that SRV is signed after all, the initial tool I used just did not support listing SRV.

Any worries regarding DNS overhead is valid but there are efforts to alleviate this https://tools.ietf.org/html/draft-bellis-dnsext-multi-qtypes-04
using multiple QTYPE DNS in a single request.

Marking this as WontFix. We've explored and decided against SRV records several times now, and have no plans to support.  Issue 22423  was one such discussion.

I'm gonna dupe against that so the (imperfect) connection is clearer.