When IPv6?

There is much fuss about IPv6. Predictions seem to be to bring it in about
2 years hence. But a search at my ISP's site returned not one match.

What are the latest guesses? Will Australia with its smaller population
have IPv4 longer? (We are just closing down DCMA cellphone networks.)

Doug.


Doug Laidlaw

On Fri, 18 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in
article <i3amd5-(E-Mail Removed)>, Doug Laidlaw wrote:

>There is much fuss about IPv6. Predictions seem to be to bring it in
>about 2 years hence. But a search at my ISP's site returned not one
>match.

IPv4 is getting scarce. As of mid-week, there were 2633858304 (2.6e9)
addresses assigned or allocated by the five RIRs (allocated = available
for sub-assignment; assigned = assigned to an end user entity), which
is 71.06 percent of the non-reserved (RFC3330) address space. A month
earlier the numbers were 2609197794 and 70.40%, while at mid-December
it was 2564407724 and 69.19%. While the five RIRs are still handing
out chunks of IPv4-land as large as a /12 (255.240.0.0, or a 1e6
addresses), this is getting pretty rare - I see a LOT more new blocks
smaller than a /19 (255.255.224.0, or 8192 addresses).

IPv6 is getting more common, but is still the rare bird. The RIRs are
handing out _huge_ IPv6 blocks (there are 6 /64s world wide, each /64
being 7e9 times as large as all of IPv4-land, and the next smallest
IPv6 block is a /48 which is 16 times wider), but there simply aren't
that many of them. Here's an interesting statistic:

RIR blocks IPv4 addresses blocks IPv6 addresses x 10e30
AFRINIC 1269 19566080 49 3.328
APNIC 14749 453129728 559 1917.837
ARIN 38863 1565847552 591 33.277
LACNIC 2088 70929920 99 10.379
RIPE 28782 524385024 1090 2608.667

Most of those IPv6 blocks in APNIC are in Japan (210), Korea (84) and
Australia (44), while in RIPE it's Germany (168), GB (129), The
Netherlands (81) and France (64) for a similar percentage.

>What are the latest guesses?

Can't help there - but there's only roughly a billion IPv4 left.
As you are interested in what's happening in Oz, you might have some
luck talking to ausregistry.net.au. Anecdotally, I heard from a
network admin in Sydney who is involved in moving a server facility.
freeing up a /24 for use elsewhere in their system. Someone else
had submitted a claim with ausregistry trying to take over the
entire /19 block (that the /24 was in) as being "unused". Like I say,
IPv4 is getting scarce.

>Will Australia with its smaller population have IPv4 longer?

Why not? Lessee, you've got four blocks (2 /20s and 2 /24s) out of
ARIN, and the rest...

[compton ~]$ zgrep AU APNIC.gz | cut -d' ' -f3 | sort | uniq -c | column
6 255.248.0.0 44 255.255.128.0 339 255.255.248.0
12 255.252.0.0 73 255.255.192.0 461 255.255.252.0
11 255.254.0.0 258 255.255.224.0 947 255.255.254.0
288 255.255.0.0 265 255.255.240.0 3170 255.255.255.0
[compton ~]$

out of APNIC - that's 34,915,840 IPv4 addresses total in 5878 blocks.
Oz also has 44 IPv6 allocations/assignments

[compton ~]$ zgrep AU IPv6.current.data.gz | cut -d' ' -f3 | sort |
uniq -c | column
2 20 2 33 5 35
24 32 2 34 9 48
[compton ~]$

and one of those small /48 is 1.208 x 10e24 which gives 5.85e16 addresses
for _each_ person in Oz (2006 population of 20.68 million). IPv4 is not
going away tomorrow, next week or even next year. Next decade might be
another story - but that's a ways out there.

>(We are just closing down DCMA cellphone networks.)

Yeah, but that's so they can offer new/improved/more-expensive stuff
to the locals. (If you think IPv4 address space is tight, go try to
get a 10 MHz wide channel in the 1-18 GHz bands used for satellite or
cell-phone services.)

Old guy

"Doug Laidlaw" <(E-Mail Removed)> wrote in message
news:i3amd5-(E-Mail Removed)...
> There is much fuss about IPv6. Predictions seem to be to bring it in
about
> 2 years hence. But a search at my ISP's site returned not one match.
>
> What are the latest guesses? Will Australia with its smaller population
> have IPv4 longer? (We are just closing down DCMA cellphone networks.)

For some of us, 3+ years AGO. ;-)

Why wait? Upgrade your systems now - because you can.

In article <fub3fa$u37$(E-Mail Removed)>, D. Stussy wrote:
> For some of us, 3+ years AGO. ;-)

For some of us, never.

> Why wait? Upgrade your systems now - because you can.

Because I have no interest in it. I disable IPV6 on every system
that I work with.

--
Roger Blake
(Subtract 10s for email.)

On 19 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in article
<(E-Mail Removed)>, Roger Blake wrote:

>D. Stussy wrote:

>> For some of us, 3+ years AGO. ;-)

>For some of us, never.

I wouldn't go so far as to say "never", but certainly the mad rush
to the future isn't effecting us yet. Several years from now, perhaps.

>> Why wait? Upgrade your systems now - because you can.

I don't control my upstream. Most people using Linux are using a
distribution somewhere close to modern (IPv6 has been included in
many distributions for over 6 years), and thus there is no need to
update anything local for that reason alone.

>Because I have no interest in it. I disable IPV6 on every system
>that I work with.

A lot of people have no direct IPv6 access, and the most common
problem is an ISP run name server[s] that ignores a 'AAAA' record
request, rather than immediately returning a NOTIMP or REFUSED
result. The result is a 5 to 15 second delay before your resolver
decides things aren't working and then tries a 'A' record request.
It's possible to set a lower timeout in /etc/resolv.conf, but this
is often a bad idea. The better solution is to swear at the idiots
who configured your DNS, and disable IPv6 on your systems. Bitching
to the ISP is usually useless, as the script-monkey you get to talk
to only knows what's in his manual, and couldn't spell IPv6 if
asked. Even if you know how to look up the SOA record, or try an
address like '(E-Mail Removed)', your chance of contacting the correct
idiot to inform that his DNS is b0rked is rather poor.

Old guy

Moe Trin wrote:

> On 19 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in
> article <(E-Mail Removed)>, Roger Blake wrote:
>
>>D. Stussy wrote:
>
>>> For some of us, 3+ years AGO. ;-)
>
>>For some of us, never.
>
> I wouldn't go so far as to say "never", but certainly the mad rush
> to the future isn't effecting us yet. Several years from now, perhaps.
>
>>> Why wait? Upgrade your systems now - because you can.
>
> I don't control my upstream. Most people using Linux are using a
> distribution somewhere close to modern (IPv6 has been included in
> many distributions for over 6 years), and thus there is no need to
> update anything local for that reason alone.
>
>>Because I have no interest in it. I disable IPV6 on every system
>>that I work with.
>
> A lot of people have no direct IPv6 access, and the most common
> problem is an ISP run name server[s] that ignores a 'AAAA' record
> request, rather than immediately returning a NOTIMP or REFUSED
> result. The result is a 5 to 15 second delay before your resolver
> decides things aren't working and then tries a 'A' record request.
> It's possible to set a lower timeout in /etc/resolv.conf, but this
> is often a bad idea. The better solution is to swear at the idiots
> who configured your DNS, and disable IPv6 on your systems. Bitching
> to the ISP is usually useless, as the script-monkey you get to talk
> to only knows what's in his manual, and couldn't spell IPv6 if
> asked. Even if you know how to look up the SOA record, or try an
> address like '(E-Mail Removed)', your chance of contacting the correct
> idiot to inform that his DNS is b0rked is rather poor.
>
> Old guy

I have ipv6 disabled at the moment. My real question was: when will I need
to turn it on again? A "host" lookup yesterday gave addresses in both
formats.

Doug.

On Sun, 20 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in
article <drgqd5-(E-Mail Removed)>, Doug Laidlaw wrote:

>Moe Trin wrote:

>> A lot of people have no direct IPv6 access, and the most common
>> problem is an ISP run name server[s] that ignores a 'AAAA' record
>> request, rather than immediately returning a NOTIMP or REFUSED
>> result. The result is a 5 to 15 second delay before your resolver
>> decides things aren't working and then tries a 'A' record request.

>I have ipv6 disabled at the moment. My real question was: when will
>I need to turn it on again? A "host" lookup yesterday gave addresses
>in both formats.

Two questions - first, are you finding you need to access any host
or service that ONLY runs on IPv6? (At least for me, I've not seen
any indication of a specific IPv6 only situation. "YMMV") And
second, if you enable IPv6, does your ISP provide you with a "real"
IPv6 address (in the range 2000:: through 4000:: minus 1, as opposed
to Link Local addresses in the E800::/64, E900::/64, FE80::/64 or
FEC0::/64 ranges). As mentioned up-thread, OZ has 44 IPv6 ranges (2
assigned to end-users, 42 allocated to providers for sub-assignment).
Is your ISP one of the lucky few?

The normal reason to disable IPv6 is b0rken DNS servers. If that is
no longer a problem for you - and you have made (or can make) any
needed IPv6 firewall rules - go for it. You _MAY_ find that IPv6
still causes some delays (a fraction of a second to make an IPv6
query and either find NXDOMAIN, or get NOERROR but not be able to
route packets "there"), but you can probably only discover this by
testing. There may also be minor slowness because the IPv6 world
isn't as well connected as IPv4 - I've done a traceroute6 from a
site in New York City, to a site in Montreal, and the packets were
routed via Japan because there was no direct route otherwise.

Old guy

Doug Laidlaw wrote:
> There is much fuss about IPv6. Predictions seem to be to bring it in about
> 2 years hence. But a search at my ISP's site returned not one match.
>
> What are the latest guesses? Will Australia with its smaller population
> have IPv4 longer? (We are just closing down DCMA cellphone networks.)
>

Don't know about the rest of the world, but here, in the Netherlands, a
lot of ISPs already have their whole infrastructure running on IPV6.
It's just not noticable for customers, since they are being provisioned
over IPV4. (compatbility stuff, etc).

In article <(E-Mail Removed)>, Moe Trin wrote:
> I wouldn't go so far as to say "never", but certainly the mad rush
> to the future isn't effecting us yet. Several years from now, perhaps.

Not for me. I've been working with "IPV4" practically since its inception
in the early 1980s and am unwilling to learn a new protocol. Hopefully
the "need" to actually use IPV6 will come after I retire from the DP
business, after which I will not need or want internet access and it
beomes a moot point. (The rest of you can deal with it as you see fit.

--
Roger Blake
(Subtract 10s for email.)

On 21 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in article
<(E-Mail Removed)>, Roger Blake wrote:

>Moe Trin wrote:

>> I wouldn't go so far as to say "never", but certainly the mad rush
>> to the future isn't effecting us yet. Several years from now, perhaps.

>Hopefully the "need" to actually use IPV6 will come after I retire
>from the DP business, after which I will not need or want internet
>access and it beomes a moot point.

Reminds me of a network tech who stated that when he retired, he was
going to walk away from town carrying a 50 foot AUI cable coiled up
on his arm, and when he found someone who asked him "where are you
going with that gray rope" he had found the place where he was going
to life thereafter.

>(The rest of you can deal with it as you see fit.

I suspect that will be harder and harder to escape. I was in for
surgery earlier this year, and while recovering enough to be able to
go home, spoke with a tech who came in to fix a problem with the
medical monitor system they had me wired up to (BP, EKG, Blood
oxygen levels, $DEITY knows what else). The monitor in the room was
connected to the nurses stations over IP.

Old guy