My multihoming solution was homegrown, and not a vendor solution. I agree that despite RFC 7084 being issued in November, 2013 there does not yet appear to be a standardized way to multihome more than one provider connection without either using PI space and BGP, or a form of NAT for IPv6 multihoming (Shim 6) on standard ISP connections. No one to my knowledge has ever actually put this RFC into action. The key part of that RFC is router notifications of zero preferred lifetime when a link fails. This signals the attached hosts to change their default gateway.

I have low margins in my smallish business, and the additional costs of PI space and BGP based ISP connections are not worth incurring just to avoid a 1 minute latency issue that might happen less than a couple of times in a month on average. Others with a bigger budget might make a different choice. Happy eyeballs is another reason not to incur costs, since a failover to IPv4 will happen in less than that minute, and of course since IPv4 sits behind a NAT, that fallover problem has been solved by almost every major vendor without the use of BGP.

I suspect the true reason for the IETF's failure on their renumbering test was the lack of implementation of things like RFC 7084. Had there been vendor support for this, the renumbering test would have gone more smoothly.

My opinion is that IPv6 needs to avoid all the hacks that we have done in IPv4 and remain pure. IPv6 brings back true end to end connections without all the hacks needed in an IPv4 enviroment, and should make things simpler with a smaller routing table. For that reason, I am opposed to IPv6 transfers. If transfers should be permitted, these transfers should be limited to PI space and it should not be permitted in PA space since that leads us down the road to much fragmented routing tables.

Albert Erdmann
Network Administrator
Paradise On Line Inc.

On Wed, 3 Apr 2019, JORDI PALET MARTINEZ via ARIN-PPML wrote:

Sorry the late answer been extremely busy for a few days and had big email 
backlog.

El 27/3/19 23:12, "[email protected] en nombre de [email protected]" 
<[email protected] en nombre de [email protected]> escribió:

   On Tue, 26 Mar 2019, JORDI PALET MARTINEZ via ARIN-PPML wrote:

   >
   > El 26/3/19 23:23, "[email protected] en nombre de 
[email protected]" <[email protected] en nombre de 
[email protected]> escribiÓ:
   >
   >    I am opposed.
   >
   >    IPv6 policies have been designed from the beginning to limit the growth 
of
   >    the global routing tables. Policies such as sparse assignment help with
   >    this goal, as well as the development of means to renumber with relative
   >    ease, compared to IPv4.  This is because more than one upstream can be
   >    advertised at the same time and in the same network.  A RFC compliant 
host
   >    will by default assign addresses in each subnet that it hears router
   >    advertisements and spread its outgoing traffic between the available
   >    upstream routers.  Unlike IPv4, we are nowhere near exhaust, and there 
is
   >
   > Have you really tried that ? Thinks aren't so easy as it may seem.

   Yes, I have done this for the last 6 years.  It works 100% on inbound
   traffic.  I agree there is a big issue with MS boxes choosing just a
   single outbound IPv6 route and never failing over, except to IPv4 because
   of happy eyeballs in the browser.

Working only for inbound traffic is not a solution.

   >
   >    no need to get into the legacy transfer issue with IPv6.  I would perfer
   >    to allow each IPv6 block assigned to a RIR to remain 100% under the
   >    control of that RIR. If transfers are possible, this fact alone can be
   >    used to defeat the trust anchor.
   >
   >    It is unclear to me what the trust anchor problem actually is, and why 
it
   >    needs to lead to the explosion of the IPv6 DFZ because of transfers.  If
   >
   > This policy will not increase the DFZ. The same number of routes you
   > have in one region will just move to another one, same or different
   > upstreams.
   >

   IPv6 was designed from the beginning to allow more than one router on a
   lan. Therefore each router can be associated with a different ISP. Hosts
   will assign an address on each network that sends a router advertisement.
   Since I am using only the ISP default route on each /64 ISP provided
   network, you are correct, the number of routes in the DFZ do not change.

   This is also the way you can easily (compared to IPv4) migrate to a new
   address block, by marking the old one as depreciated. At that point, old
   connections are continued, but all new ones happen on the new address
   block.

You mean renumbering. We had a big IETF effort to try that. Is not that easy. 
It means somebody need to change, for example, DNS. This means that you can't 
have multihoming with automatic failover, load sharing/balancing, etc., not 
that easy.

   >    there is an issue of ARIN policy regarding trust anchors compared to 
other
   >    RIR's, this policy should instead be addressed instead of allowing
   >    transfers as a work around to a bad ARIN policy.
   >
   >    Ideally, IPv6 blocks should be obtained from the upstream ISP/LIR and 
they
   >    should be routed to the default route, with only one route per ISP/LIR.
   >
   > Ideally yes, one router per ISP/LIR, but not all follow that because
   > "traffic engineering".

   True, but usually all customers in a given area use that same set of
   routes, and this does not require any route involvement at the ISP.

   >
   > If I understand correctly, you're saying that we must not have IPv6 PI?
   >

   No, I am not saying that at all.  I am pointing out that in IPv6 that
   multihome and failover should work over standard ISP connections, without
   any involvement with the upstream ISP. This also allows failover without
   having to obtain IPv6 PI space, using only the upstream provided IPv6
   addresses. If you have a site that needs more bandwidth than a standard
   connection can provide, you can choose to use two or more different
   providers to provide the required bandwidth and also obtain failover
   without any additional costs. Each host will obtain an IPv6 address from
   every available router. In the case of MS windows, it will only use one of
   the provided IPv6 /64's for outbound connections, but which one is chosen
   is quite random and tends to be close to evenly split between each of them
   in a group of MS windows machines.  I wish that MS would read the RFCs and
   split their outbound bandwidth use between all available connections like
   Linux does.

   >    Since the "normal" site assignment is a /48, unlike IPv4, there is no
   >    shortage of address space for any use without involvement of ARIN or 
other
   >    RIR.  If one needs to be multihomed, each host can have an address from
   >    each available upstream, providing availability to each host from more
   >    than one network.
   >
   > Have you tried that? Hosts don't multihome today with IPv6 with multiple
   > prefixes from different upstreams. You really need IPv6 PI.
   >

   I have, and have been doing it for the last 6 years.  It does require DNS
   tricks.  I have a script running that removes the AAAA record for any
   upstream that is not available, and replaces it when it returns. Of course
   that record uses a TTL of 60. It simply pings each circuit's gateway
   address to determine the up/down status of that connection once a minute,
   and adds/removes the DNS records for the IP address for that connection
   assigned to each host on the network that has either failed or returned.
   While not instant or perfect, the script is cheap and requires no
   additional hardware or costs.

Not instant means is not usable. If it means having a low TTL is bad for being 
able to cache the RR, etc.

I can't agree this is the way at all.

   >    Albert Erdmann
   >    Network Administrator
   >    Paradise On Line Inc.
   >
   >    On Tue, 26 Mar 2019, ARIN wrote:
   >
   >    > On 21 March 2019, the ARIN Advisory Council (AC) accepted 
"ARIN-prop-263:
   >    > Allow Inter-regional IPv6 Resource Transfers" as a Draft Policy.
   >    >
   >    > The Draft Policy text is below and can be found at:
   >    > https://www.arin.net/participate/policy/drafts/2019_4/
   >    >
   >    > You are encouraged to discuss all Draft Policies on PPML. The AC will
   >    > evaluate the discussion in order to assess the conformance of this 
draft
   >    > policy with ARIN's Principles of Internet number resource policy as 
stated in
   >    > the Policy Development Process (PDP). Specifically, these principles 
are:
   >    >
   >    > * Enabling Fair and Impartial Number Resource Administration
   >    > * Technically Sound
   >    > * Supported by the Community
   >    >
   >    > The PDP can be found at:
   >    > https://www.arin.net/participate/policy/pdp/
   >    >
   >    > Draft Policies and Proposals under discussion can be found at:
   >    > https://www.arin.net/participate/policy/drafts/
   >    >
   >    > Regards,
   >    >
   >    > Sean Hopkins
   >    > Policy Analyst
   >    > American Registry for Internet Numbers (ARIN)
   >    >
   >    >
   >    >
   >    > Draft Policy ARIN-2019-4: Allow Inter-regional IPv6 Resource Transfers
   >    >
   >    > Problem Statement:
   >    >
   >    > There is an operational need to allow RIR transfers of IPv6 resources 
between
   >    > RIRs with an equivalent transfer policy. ARINâ•˙s RPKI Trust Anchor 
(TA) is
   >    > measurably less widely deployed than TAs from other RIRs. As a 
consequence,
   >    > RPKI ROAs published through ARIN offer less value. Operators seeking 
to
   >    > extract the most value from their investment in IPv6 would benefit 
from the
   >    > ability to transfer IPv6 resources to RIRs with more widely deployed 
RPKI
   >    > Trust Anchors.
   >    >
   >    > Policy Statement:
   >    >
   >    > Change the first sentence in section 8.4 from:
   >    >
   >    > ╲Inter-regional transfers of IPv4 number resources and ASNs may 
take place
   >    > only via RIRs who agree to the transfer and share reciprocal, 
compatible
   >    > needs-based policies.╡
   >    >
   >    > To:
   >    >
   >    > ╲Inter-regional transfers of Internet number resources may take 
place only
   >    > via RIRs who agree to the transfer and share reciprocal, compatible
   >    > needs-based policies.╡
   >    >
   >    > Comments:
   >    >
   >    > Timetable for implementation: Immediate
   >    > _______________________________________________
   >    > ARIN-PPML
   >    > You are receiving this message because you are subscribed to
   >    > the ARIN Public Policy Mailing List ([email protected]).
   >    > Unsubscribe or manage your mailing list subscription at:
   >    > https://lists.arin.net/mailman/listinfo/arin-ppml
   >    > Please contact [email protected] if you experience any issues.
   >    >_______________________________________________
   >    ARIN-PPML
   >    You are receiving this message because you are subscribed to
   >    the ARIN Public Policy Mailing List ([email protected]).
   >    Unsubscribe or manage your mailing list subscription at:
   >    https://lists.arin.net/mailman/listinfo/arin-ppml
   >    Please contact [email protected] if you experience any issues.
   >
   >
   >
   >
   > **********************************************
   > IPv4 is over
   > Are you ready for the new Internet ?
   > http://www.theipv6company.com
   > The IPv6 Company
   >
   > This electronic message contains information which may be privileged or 
confidential. The information is intended to be for the exclusive use of the 
individual(s) named above and further non-explicilty authorized disclosure, 
copying, distribution or use of the contents of this information, even if 
partially, including attached files, is strictly prohibited and will be considered 
a criminal offense. If you are not the intended recipient be aware that any 
disclosure, copying, distribution or use of the contents of this information, even 
if partially, including attached files, is strictly prohibited, will be considered 
a criminal offense, so you must reply to the original sender to inform about this 
communication and delete it.
   >
   >
   >
   > _______________________________________________
   > ARIN-PPML
   > You are receiving this message because you are subscribed to
   > the ARIN Public Policy Mailing List ([email protected]).
   > Unsubscribe or manage your mailing list subscription at:
   > https://lists.arin.net/mailman/listinfo/arin-ppml
   > Please contact [email protected] if you experience any issues.
   >_______________________________________________
   ARIN-PPML
   You are receiving this message because you are subscribed to
   the ARIN Public Policy Mailing List ([email protected]).
   Unsubscribe or manage your mailing list subscription at:
   https://lists.arin.net/mailman/listinfo/arin-ppml
   Please contact [email protected] if you experience any issues.




**********************************************
IPv4 is over
Are you ready for the new Internet ?
http://www.theipv6company.com
The IPv6 Company

This electronic message contains information which may be privileged or 
confidential. The information is intended to be for the exclusive use of the 
individual(s) named above and further non-explicilty authorized disclosure, 
copying, distribution or use of the contents of this information, even if 
partially, including attached files, is strictly prohibited and will be 
considered a criminal offense. If you are not the intended recipient be aware 
that any disclosure, copying, distribution or use of the contents of this 
information, even if partially, including attached files, is strictly 
prohibited, will be considered a criminal offense, so you must reply to the 
original sender to inform about this communication and delete it.



_______________________________________________
ARIN-PPML
You are receiving this message because you are subscribed to
the ARIN Public Policy Mailing List ([email protected]).
Unsubscribe or manage your mailing list subscription at:
https://lists.arin.net/mailman/listinfo/arin-ppml
Please contact [email protected] if you experience any issues.
_______________________________________________
ARIN-PPML
You are receiving this message because you are subscribed to
the ARIN Public Policy Mailing List ([email protected]).
Unsubscribe or manage your mailing list subscription at:
https://lists.arin.net/mailman/listinfo/arin-ppml
Please contact [email protected] if you experience any issues.

Reply via email to