Re: HTTPS RR side of things
Daniel Stenberg via curl-library <[email protected]>
| Newsgroups | gmane.comp.web.curl.library |
|---|---|
| Message-ID | <[email protected]> |
On Fri, 21 Feb 2025, Niall O'Reilly wrote: > I wonder whether you've considered res_nquery(), the non-blocking successor > to res_query()? I see it in glibc; I haven't looked in Apple or BSD land. > Building without having to install an extra package (such as c-ares) might > be attractive. Perhaps there are disadvantages? Is it really non-blocking? The docs I've read does not seem to say so. There are (other) significant drawbacks with using this API: primarily that it is limited to just a small subset of platforms. It does not even exist on Windows or macOS and it's broken on Android etc. We would have to use a wide range of different native APIs to cover just the most commonly used OSes - way more work than just making sure that a solid DNS library like c-ares is used instead. Mozilla took the "native API" route, and Valentin Gosu's talk from this year's FOSDEM provides reasons enough for me to avoid selecting this approach: "getaddrinfo sucks, everything else is much worse" The video recording is here: https://video.fosdem.org/2025/h2213/fosdem-2025-4229-getaddrinfo-sucks-everything-else-is-much-worse.mp4 -- / daniel.haxx.se || https://rock-solid.curl.dev -- Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.html