Am 02/11/13 23:19, schrieb Mark Felder: > > On Nov 2, 2013, at 3:27 PM, Adrian Chadd <adrian_at_freebsd.org> wrote: > >> A lot of HTTP infrastructure lives on anycast DNS, HTTP redirects and >> geoip records. Saying it's broken and not feasible is nonsense. > > More specifically what I was referring to was the fact that traditionally HTTP failover with round-robin A records is very unreliable; every client can act differently. You really need to be doing anycast as well to ensure those records are always available which adds additional architecture complexity that the project may not have the resources to throw at. GeoDNS also adds a layer of complexity, but as it turns out there are members of the project with extensive experience running it. SRV would give us very simple, cheap, reliable failover. It seems we do have some blockers, though. > > The good news is that we fully control the client. Hopefully we can just work around these issues. Hi, I like green better than purple! Everybody is just rationalizing their positions with great sophistication. I think the new behavior is a POLA violation. I was pretty annoyed when i could not browse the given URL. And this will probably result in lots of wasted time for admins, becuase we are doing it differently from everybody else. Everybody else is doing pretty fine with the status quo. There should be a better reason than the listed to "roll our own". - With this threshold, we could also start to replace (smtp, http, dns, pam, posix ....). [*cough* systemd] I am not a FreeBSD developer, so see it as the opinion of a non-paying consumer :) Thanks for pkg-ng, its awesome so far!Received on Sun Nov 03 2013 - 01:00:36 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:43 UTC