Simpler DNS interface needed

[schveiguy]

Simpler DNS interface needed

[schveiguy]

Also, I think it would be nice to put the wrapper function inside the dns module. Having to use the driver directly seems against the spirit of this library.

[MartinNowak]

Hence it's private. I tried a couple of things, but getaddrinfo is really only designed to open a socket. On the good side, it can transparently handle IPv4 vs. IPv6 and multiple IPs if used correctly.
DNS isn't a hostname -> addr mapping. And what would you do with a list of addresses of different families?
My takeaway was to use the C resolver for sockets, but leave DNS to some other library. A real DNS library would also support other queries like MX, TXT, et.al.
This reasoning should be properly documented somewhere.

[MartinNowak]

What's the use-case for a manual DNS resolution of the hostname @schveiguy?
Resolving names once on program startup is somewhat of an anti-pattern as it defies certain server-side failover/loadbalancer mechanisms (hence TTL in caching in recursive resolvers).
Also sockets can be used over longer times, so repeated lookup shouldn't be too much of a problem, and most ppl. have caching resolvers nearby anyhow.

[schveiguy]

Hence it's private

Huh? I was able to use it here: https://github.com/schveiguy/httpiopipe/blob/master/source/iopipe/http.d#L591

DNS isn't a hostname -> addr mapping

I have no idea what you mean by this statement. What other service would I use to open a socket to a named host?

I think there's a misunderstanding here. I'm not complaining about the service or the usage of getaddrinfo, just that I have to use the driver directly (which is completely different from most of std.io), and that the API requires me to write my code in a node.js callback style of code.

[schveiguy]

What's the use-case for a manual DNS resolution of the hostname @schveiguy?
Resolving names once on program startup is somewhat of an anti-pattern as it defies certain server-side failover/loadbalancer mechanisms (hence TTL in caching in recursive resolvers).
Also sockets can be used over longer times, so repeated lookup shouldn't be too much of a problem, and most ppl. have caching resolvers nearby anyhow.

Literally, I have no idea why you wrote any of this. I really do feel you are reading some other meaning than what I wanted to say.