Re: TCP v6 and LuaSocket

[Daniel Silverstone <dsilvers@...>]

On Tue, Mar 22, 2011 at 12:23:15PM +0100, Axel Kittenberger wrote:
> Well, ipv6 works different than ipv4 in that every machine always has a ipv6
> address for the ipv6 local connect network. The knowledge that there arent
> further routes to the rest of the ipv6 is kinda of a different place, and
> there isnt a fast fail. 

The "local connect network" I assume you mean the link-local addressing
(fe80::) and yes, while that is always present, there's no routes presented for
global routing with that and as such, the local host should generate ICMP
messages saying the network/host is unreachable immediately (thus failing the
connect() call).

> And I can assure you that my router/firewall does not do any v6 advertisment.
> Still OS X or now also new Linux machines per default will go into long
> timeout trying to connect to a v6.

I do not experience this on any of our networks with modern Linux boxes.  They
all get their link-local address and correctly instantly fall back to ipv4
without issue.

> Yes, we'll have a real v6 network access in the near to medium future, where
> these issues will vanish automagically.  But until then, there are problems
> due to first v6 connect, and usually the solution is to configure the hosts
> to disable the v6 stack.

I'm still finding this claim incredulous but I do agree with the overall
requirement to be able to turn it off and support the idea of luasocket having
connect/resolve options for this.

D.

-- 
Daniel Silverstone                         http://www.digital-scurf.org/
PGP mail accepted and encouraged.            Key Id: 3CCE BABE 206C 3B69