[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: why I'm not implementing IPv6



On 21 Nov 2002, D. J. Bernstein wrote:
> In short: the same reasons I didn't implement OSI. In detail:
> 
>    http://cr.yp.to/djbdns/ipv6mess.html

Speaking of which, I'll add my private (but smaller, more constrained) 
rant.

IMO, it was a serious mistake to make the system libraries sort AAAA DNS
replies before A records (btw: I'm not aware of any implementation where
you could actually specify the order) when using getaddrinfo and
AF_UNSPEC.

This is because now people are afraid of deploying v6-enabled software
because v6 connectivity is not really there, and contacting dual-stack
servers is only likely to cause problems when done over v6.  Only when
explicitly configured, v6 would be used over v4, until some more or less 
"magic moment" (at different times in different regions etc.)

If this would have been different, we migh have a significantly higher
number of v6-enabled applications etc. enabled in the mainstream
distributions.

Further, this will affect the transition scenarios.  I surely wouldn't
like regular customers' getting v6-enabled yet with this bad connectivity.

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords