TID: FMail-W32 2.1.3.7-B20170919
RFC-X-No-Archive: Yes
TZUTC: 0100
CHRS: CP850 2
MSGID: 2:280/5555 622cb026
REPLY: 2:240/1120.29 473fbe83
Hello Tim,

On Saturday March 12 2022 02:11, you wrote to Wilfred van Velzen:

 TS>>> I have changed WP to use the first address returned by the OS.
 TS>>> We have to test if that works as expected.

 WvV>> And if the first fails or times out, it should move on to the
 WvV>> next (that's what binkd does)...

 TS> Yes, that is the idea. But this is a little more work and thus not
 TS> something I do right now. Again, this currently makes preferring IPv4
 TS> over IPv6 probably the better choice, as it addresses exactly the
 TS> scenario where the fallback would be needed.

A fallback is in order when there is more than one IP adress to choose from
and the attempt to connect with the first choice fails for one reason or
another. This is not related to IPv6 vs IPv4 per s‚. It could be that the host
presents several IPv4 addresses but no IPv6 address. Or the other way around,
it presents more than one IPv6 address but no IPv4 address. Or it could
produce a mix of IPv4 and IPv6 addresses. But let us focus on the case that
the choice is between IPv6 and IPv4.

Your conjecture that IPv4 is the better choice instead of the choice presented
by the OS in case there is no fallback mechanism is based on the assumption
that in general IPv4 has a better chance of resulting in a connection than
IPv6. As we discussed before, that may have been a valid assumption ten years
ago, but it certainly does not hold today. Yes, one may still run into a
situation where an IPv6 connection fails and an IPv4 connection can be
established. But one may just as well run into the reverse situation. I have
no personal experience with the latter situation. My system always tries the
OS choice first, which (nearly) always is IPv6. So I do not know if it ever
happens that an IPv4 connect fails and a subsequent IPv6 succeeds. My binkd
does not try to outsmart the OS in choosing between IPv4 and IPv6. But I have
no reason to think that the one will happen more often than the other. In this
case I do not think it is a good idea to try to outsmart Windows...

 TS> I still keep thinking (and nothing brought forward so far has been a
 TS> valid argument against it) that in practice there are usually only
 TS> benefits and no drawbacks of preferring IPv4 over IPv6 when both are
 TS> available.

I already mentioned the situation that the Winpoint user is on a DS-Lite
connection and the BOSS node is Dual Stack. I also mentioned the situation
that the BOSS node is on a DS-LIite connection but presents an (invalid) IPv4
address nevertheless.

 TS> On the other side, there usually exists absolutely no benefit for the
 TS> user in choosing IPv6 in that scenario.

Maybe not from the POV of the user, but for the internet as a whole choosing
IPv4 where IPv6 is possible is detrimental regarding the transition from IPv4
to IPv6. That transition is unavoidable and the faster it is completed, the
better.


Cheers, Michiel

--- GoldED+/W32-MSVC 1.1.5-b20170303
 * Origin: http://www.vlist.eu (2:280/5555)
SEEN-BY: 15/0 106/201 124/5016 129/331 153/757 7715 203/0 218/700
SEEN-BY: 221/0 229/110 317 426 428 664 700 266/512 280/464 5003 5555
SEEN-BY: 282/1038 292/854 8125 301/1 310/31 317/3 320/219 341/234
SEEN-BY: 396/45 460/58 712/848 770/1 2452/250
PATH: 280/5555 464 712/848 229/426