Split(".www.miek.nl.") -> []string{"", "www", "miek", "nl"},

[miekg]

This might be a bug (or not). It's is basically an illegal domain name that you then split...

[andrewtj]

Change the interface, add a new func or make a note of it for a someday-maybe dns2?

[miekg]

We could return []{0} for this... but yeah returning an error is better...

[andrewtj]

Maybe a simple breaking change log could make things like this easier. A low effort way to do it would be to just put '#breaking` or similar in commit messages and then extract into the readme a list of commits that break things. This would be even more useful for changes that silently break things like #281 seems to have done to rrda (see fcambus/rrda#11).

[miekg]

[ Quoting notifications@github.com in "Re: [dns] Split(".www.miek.nl.") ->..." ]

Maybe a simple breaking change log could make things like this easier. A low effort way to do it would be to just put '#breaking` or similar in commit messages and then extract into the readme a list of commits that break things. This would be even more useful for changes that silently break things like #281 seems to have done to rrda (see fcambus/rrda#11).

I agree. Should still be careful, but bugs need to be fixed even though that could
break backwards compat.

I also like the '#breaking' idea.

/Miek

Miek Gieben

[andrewtj]

Sketched up a way it could work.

[miekg]

Coming back to this specific issue and looking through other label function, such as CountLabel for instance, they don't return a value plus error. So either we break them all, or we assume and document that in the input should be a properly formatted dns name/label. I'm leaning towards the latter (garbage in, garbage out)

[andrewtj]

That sounds okay. If there's demand in the future error returning variants could always go in dnsutil.

[miekg]

[ Quoting notifications@github.com in "Re: [dns] Split(".www.miek.nl.") ->..." ]

That sounds okay. If there's demand in the future error returning variants could always go in dnsutil.

Ok, let's agree to that. They are some prinln("deprecated") littered in the
source and at some point all of these functions/methods should be deprecated
(pref. in one big bang release)

/Miek

Miek Gieben