On 20 Feb 2025, at 21:32, Daniel Stenberg made some helpful comments, which I appreciate very much; he wrote:
> Isn't this pretty much what is already being attempted by getdns, libunbound, > c-ares and the likes? Yes, but not exactly, I think. Using any of these is ideal iff it is acceptable for the application to ignore "local policy" (nsswitch, ipv6 address selection). ISTR mail from you quite a while ago that made me appreciate (after a while: I might be a slow learner) that getaddrinfo() isn't such an abomination as I used to think it was. 8-) Thanks for that. > But yes, a worthy project it sounds like! Thanks for the encouragement! > Writing a getaddrinfo drop-in replacement is a task that has been attempted > and failed countless number of times by competent developers. > > So even without adding anything extra, doing this "right" is hard. Yup. > Happy Eyeballs v2 is enough of a challenge as it for example does not work > with getaddrinfo due to its limited API. Yup. > It'll be interesting to see what v3 changes in the mix. Yup again. I guess you have read, or soon will read, the draft, so I don't need to comment any more extensively. > I suspect the real complications will probably end up in the layers outside > of pure DNS though. Fourth yup. Thanks /Niall -- Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.html
