[6bone] IPv6 application porting feedback solicited
Stig.Venaas at uninett.no
Fri Jan 16 06:12:03 PST 2004
On Fri, Jan 16, 2004 at 12:04:30PM +0200, Pekka Savola wrote:
> If possible, please send feedback within a week. Thanks.
I have a few items.
1. In 5.1 on presentation format, it talks about semi-colon where it
should be colon. Quote:
A particular problem with IP address parsers comes when the input
is actually a combination of IP address and port. With IPv4, these
are often coupled with a semi-colon, like "192.0.2.1:80". However,
such an approach would be ambiguous with IPv6, as semi-colons are
already used to structure the address.
2. I think it's too strong to say that it's bad practice to add AAAA
record in DNS before all services support IPv6. There are many cases
where one wants to start supporting IPv6 for some services, but are
not able to IPv6 enable all. All client applications should be able
to switch to v4 if v6 fails, and it shouldn't cause noticable delay
for the user, unless there are some broken pieces in the network
mishandling AAAA queries or responses. I agree it should be avoided,
but to me "bad practice" sounds too strong.
3. Bind behaviour
The document should perhaps say something about differences in bind
behaviour. The main issue is whether you can first bind to v6 wild
card address, and then the v4. It can be a pain to write applications
that cope with this. In most cases, use of IPV6_V6ONLY should help,
but unfortunately it doesn't say anywhere that it should affect bind
behaviour. I think it's logical that it should though. The reason the
v4 wild card bind fails on some systems, is that the v4 space is
embedded into v6 space when using mapped addresses.
More information about the 6bone