Relative performance of IPv6 and IPv4 for current users

[Bill Owens <owens@nysernet.org>]

There's a small debate going on within the Apple IPv6 mailing list 
over whether it is wise to have an application prefer v6 
connectivity, when it is available, over v4. The argument against 
preferring to use v6 runs along the lines of "most people have only 
v4 connectivity, and for those few that have v6, the performance of 
v4 is usually better".

Although it is hardly a scientific way to conduct a survey, I'm 
curious what the members of this group think of that assertion. 
Personally, I find that my v6 connectivity is either just about as 
good as v4, or very much worse, and the difference is whether the 
other end is connected natively or through some variety of tunnel. 
But I don't have a good feel for what percentage of v6-accessible 
sites fall into each category.

There is also a sub-argument about 6to4, and whether it is sensible 
to enable 6to4 with anycast relay support given the dearth of relay 
routers. There seems to be some agreement that in the abscence of a 
local v6 router, allowing the machine to configure itself as a 6to4 
router without a relay is a reasonable compromise. I think that I can 
agree with that, if those v6 connection attempts which fail without 
the default route, do so rapidly enough to avoid delaying the 
eventual v4 connection. Would anyone with more 6to4 experience care 
to comment on that idea?

Bill.