Lukas Tribus

There is not good default. But that doesn't mean we cannot improve the situation. Because there is no good default, I suggest we warn/deprecate and and some point reject configurations...

@bedis copying my suggestion over from: https://www.mail-archive.com/[email protected]/msg38848.html > Are you suggesting that we should use IPv4 only unless IPv6 is set, and > that under no circumstance we switch from...

Yes, I like the global option too. Do you also want to deprecated a configuration without any resolve-prefer, per server exclusive family or global family, so that at the end...

A only slightly related question: why don't we close idle HTTP sessions on the old worker gracefully as soon as we get the signal? GOAWAY in HTTP/2, ssl_shutdown in H1...

This would introduce a lot of complexity for load-balancing, health-checks, etc which is in my opinion not worth the effort. > To me, this has the disadvantage as external monitoring...

> My hope was, that decreasing `tcp_fin_timeout` No, fin timeout is not about TIME_WAIT sockets. You could lower `tcp_max_tw_buckets`, but I doubt that is really a good idea (it will...

As already explained by Willy, this is not what the haproxy cache will do. It's a simple cache for things like the favicon. Haproxy is not nginx or varnish. You...

I have to say I disagree. I'm not sure I understand the problem in the first place about navigating a large file/text, why is it difficult to navigate a bigger...

> The current default value of 3 for `MAX_SESS_STKCTR` is quite low while dealing with a complex setup. Please elaborate and explain your exact use case with a concrete example...

@vuolter @GammaC0de as mentioned in #3250, requests instead of pycurl would be great for a number of reasons (pure python dependency, no gnutls issues), however not at the cost of...