From:  Daniel Stenberg <daniel@haxx.se>
Date:  01 Nov 2018 18:36:34 Hong Kong Time
Newsgroup:  news.mozilla.org/mozilla.dev.tech.network
Subject:  

Re: does firefox failover on DNS Round Robin?

NNTP-Posting-Host:  63.245.210.105

On Thu, 1 Nov 2018, Klaus Mueller wrote:

> I realised that browsers do failover on alternative entries on DNS Round 
> Robin. Also others noticed this, but [have found it very hard to get clear 
> confirmation that they are supposed to.][article]

This is how Internet clients in general is supposed to work since the dawn of 
time really.

You ask for all the addresses of a host, then iterate over them and if one 
fails, you continue to the next. Ideally you try IPv4 and IPv6 addresses in 
parallel, so called happy eyeballs, and stick to the one that connects first.

Note also that getaddrinfo() doesn't typically return the addresses in a 
random order so this list of adddresses is often not as "round robin" as you'd 
think. [1]

> So my queston is: is this behavior a firefox feature?

Firefox works this way too, yes, and it does happy eyeballs.

> Is there some documentation, standard or RFC?

Already RFC 1034/1035 (November 1987) recommends/discusses returning multiple 
records for a name. I don't think trying all/several addresses is mandated by 
any standard, but why wouldn't a client try the next if the first one fails?

I'm pretty sure I first learnt to do this reading Steven's network books in 
the 90s.

[1] = 
https://daniel.haxx.se/blog/2012/01/03/getaddrinfo-with-round-robin-dns-and-happy-eyeballs/

-- 

  / daniel.haxx.se