08:39 <flip214> luis: one minor problem: if I specify "0.0.0.0" or "::" to listen on, that address can't be connected to... we'd need a manual override to translate that to eg. 127.1 resp. ::1

08:40 <flip214> or is the specification/convention that only specific, set up IP addresses may be used here?

08:46 <flip214> I guess I shouldn't use usocket in swank... so I need to manually convert the octets back into an IP address

10:30 <luis> flip214: not sure what you mean by manual override

10:30 <luis> flip214: btw, you're pushing things to https://github.com/slime/slime/pull/456 because you're used the master branch for that PR

10:30 <luis> *you've

10:32 <flip214> luis: oh, ouch

10:32 <flip214> thanks for the notice

10:32 <flip214> so I should close that PR I guess

11:11 <pjb> flip214: 0.0.0.0 means listen to all interfaces. Of course, you cannot connect to it, you must choose the address of one specifikc interface.

11:24 <flip214> pjb: yeah, that's exactly the problem. the server may specify 0.0.0.0, but that doesn't help clients to connect!

11:29 <pjb> flip214: ifconfig -a

11:30 <flip214> pjb: doesn't exist on windows, on linux it becomes more and more rare, on embedded very improbable, ...

11:30 <flip214> and then parsing the output etc..... well, never mind.

11:33 <pjb> flip214: on ms-windows, it's ipconfig

11:34 <pjb> flip214: and if it's a problem, the server can get the IP addresses of the interfaces and report them.

11:34 <pjb> (write a patch).

12:44 <luis> That's what we've been discussing. SWANK will announce the IP address, right?

12:45 <luis> flip214: that's the point of the local-address interface, is it not?

12:53 <flip214> luis: yes. but if swank binds to 0.0.0.0 or ::, there _is_ no IP address to announce!

12:58 <pjb> Yes, swank issues a message with the local address.

12:58 <pjb> oh!

12:59 <pjb> then it would have to test for 0.0.0.0 (and similar in ipv6), and instead get the list of interfaces and the list of addresses for each interface.

13:03 <luis> Oh, right, 0.0.0.0 means listens on all interfaces. Forgot about that

13:03 <luis> So, hmm, don't listen on all interfaces. :P Seems like a bad idea anyway.

13:15 <jackdaniel> what was your ip again? :-)

13:23 <luis> jackdaniel: I recently learned that my ISP (like many others) assigns more than 4 sextillion addresses to each customer (256 IPv6 subnets)

13:23 <jackdaniel> and you listen on all of them with ::? :D

13:24 <jackdaniel> jokes aside, that sounds quite excesive

13:24 <luis> It's the current RFC 6177 recommendation. They used to recommend more.

13:45 <pjb> this let you avoid NAT…

13:46 <pjb> you can have a single IPv6 address, and assign addresses to each of your employees, and let them assign addresses to each of their devices.

13:46 <jackdaniel> I think that if each client had 9999999999 ip addresses, they still should be able to avoid NAT ;-)

13:46 <pjb> 128-bit…

13:46 <jackdaniel> but yeah, I get the motivation - make everything addressable

13:59 <luis> Yeah, a single 128-bit subnet (18 quintillions) seems like it would be enough, but the RFC mentions that having multiple subnets would be useful (don't remember why)

13:59 <jackdaniel> sub1: light bulbs (high security), sub2: the computer (open as internet explorer)