[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [GLOBAL-V6] PI space under IPv6
Uchenna N Ibekwe wrote:
[..]
> - Since we already have most of the PI space reserved under
> IPv4, it would be easier to continue to use the allocated IPv4 space as
> PI under the translated IPv6 numbering scheme.
With 'translated IPv6 numbering scheme', do you mean:
- IPv4 mapped: eg ::a.b.c.d
- IPv4 compat: eg :ffff:a.b.c.d
- or 6to4: eg 2002:aabb:ccdd::/48
If you mean the first two (mapped & compat); those are impossible to use
for this purpose; and actually for any other purpose they are also quite
useless as they create more problems, annoyances and pitfalls than they
are worth. One should in effect never see them being used, if one wants
to present an IPv4 address to a user, directly use a.b.c.d. On the wire
they can't be used anyway. Protocols/implementations that do are flawed
per definition.
As for 6to4, that is a huge problem. Mostly because there are only a
limited amount of anycast relays. One will thus be thrown at the mercy
of those anycasts to function. Also debugging a 6to4 problem is
unfeasible because of anycast and a lot of misconceptions on how to
properly configure a 6to4 setup.
> - Routing tables as a result will be kept within a manageable size.
No, the IPv6 routing table will include the IPv6 routing table.
The IPv6 routing table is already very well managed in size: ~800
prefixes in IPv6 versus ~200.000 in IPv4.
> - No renumbering will be required by end-users, enhancing
> migration from IPv4 to IPv6.
No renumbering!? If you want IPv6 applications to talk to other IPv6
applications you need to upgrade the applications and you will also need
to have a network supporting this addressing. It is not renumbering, but
in effect numbering, for the first time. Easily solved with RA's and
DHCPv6 of course.
> - Multi-homed users or entities will not cause delays for ISPs
> migrating to IPv6.
Which delay? And how does an endsite cause a delay for an ISP starting
to do IPv6 (note my avoidance of 'migration', as without IPv4 you are
not connected to the internet as google does not work ;) There are only
a minimal amount of applications that are IPv6 only and especially which
ones require IPv6 to function. There is, still, no critical mass and
IPv6 won't be a flag day.
With Vista this will change a bit, but still, not a lot of apps do IPv6
per default on the Windows platforms.
Also, quite a number of ISP's are already IPv6 ready internally, the
link to the endsites though is the missing part.
> This is something to think about, just an idea.
Unfortunately it won't work ;) 6to4 is a partial possibility, but due to
the the inherent brokenness of 6to4 it is not a viable solution.
If an entity requires address space they should simply get it from their
RIR. If the RIR does not have a plan for PI, then fight for it to get
it. Address Space has nothing to do with routing and RIR's are there to
provide Address Space. It is not called "Routing Space" what you get
from them.
Greets,
Jeroen
Attachment:
signature.asc
Description: OpenPGP digital signature
