[nycbug-talk] IPv6 Followup

[Alex Pilosov]

On Mon, 12 Nov 2007, Miles Nordin wrote:

> >>>>> "ap" == Alex Pilosov <alex at pilosoft.com> writes:
> 
>     ap> <snip> trust me, they exist now for v4, and nobody cares.
> 
> uh yeah, but the point is the only people remedial enough to want them
> can't use them because these people have a single dynamic IP.  Nobody
> cares because the people who can run them have better ways to do the
> same thing.  And since each box would eat an IP, I don't have that kind
> of v4 space to burn, so I wouldn't run them myself even if I didn't
> dislike them.
setting up a port 80 mapping takes a minute. how many webservers do you
want in your apartment?

> with v6, this spurned class of device gets a crack at a newer,
> less-sophisticated, and larger audience.
> 
>     ap> ips will not be global and static. more myths. you will not
>     ap> have a portable and permanently assigned IP address.
> 
>     ap> You cannot have permanent and portable v4 or v6 address.
> 
> Current jabber servers don't need PI-portable space to run, either.
> That's what DNS is for.
then what's wrong with dyndns for v4?

>     ap> If you want static IP (non-portable) you can do it just as
>     ap> well with v4 as with v6. No real difference.
> 
> yes good point.  Just having static v4 addresses like your ADSL,
> instead of dynamic, for typical home accounts would permit lots of
> things we can't do now.  but:
> 
>  (1) there's a question, which box gets the static IP?  Generally it's
>      some SOHO router.  With v6, the box on which gaim is running can
>      have a static global IP.

> 
>      Yes, I know, the IP changes when you change ISP's.  but not
>      unexpectedly and not daily/weekly.
i had same static ip on timed warner for 2+ years. ymmv.

>  (2) the hope is, aside from DoCoMo, most ISP's won't bother forcing
>      end systems onto dynamic addresses.  It's hard because:
> 
>      (a) it's not working super well to have an address changing
>          ~daily with the stateless autoconfig mechanism.  They could
>          maybe set pltime very low, though.
> 
>      (b) because now that the prefix on the wire is changing it'll
>          screw up noticeable things like local file and printer
>          sharing when your prefix changes.  There's some support for
>          keeping alive old TCP circuits on a ``deprecated'' prefix if
>          they leave vltime set to something long, but I bet it won't
>          work well.
> 
> yeah, you're right, IPv6 is only _needed_ for (1).  But I started out
> saying ``here are some reasons big ISP's may want to keep IPv6 out of
> homes as long as possible.'' (2) is such a reason---if they can't get
> their dynamic address garbage to work, they might not like giving any
> kind of static address, even a v6 one.
> 
> I hadn't considered they might give out dynamic v6 prefixes.  I guess
> that's worth worrying about.
why wouldn't they?

there's no fundamental difference between v4 and v6. stop trying to 
pretend there is - those are myths. its just more ips.

-alex