From owen@delong.com Mon Dec 11 11:11:59 2023 Return-Path: X-Original-To: v6ops@ietfa.amsl.com Delivered-To: v6ops@ietfa.amsl.com Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id EA8DFC151067; Mon, 11 Dec 2023 11:11:59 -0800 (PST) X-Virus-Scanned: amavisd-new at amsl.com X-Spam-Flag: NO X-Spam-Score: -2.095 X-Spam-Level: X-Spam-Status: No, score=-2.095 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, LOTS_OF_MONEY=0.001, RCVD_IN_ZEN_BLOCKED_OPENDNS=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_MONEY_PERCENT=0.01, T_SCC_BODY_TEXT_LINE=-0.01, URIBL_DBL_BLOCKED_OPENDNS=0.001, URIBL_ZEN_BLOCKED_OPENDNS=0.001] autolearn=ham autolearn_force=no Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=delong.com Received: from mail.ietf.org ([50.223.129.194]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id VFZLXV_Q8BKR; Mon, 11 Dec 2023 11:11:55 -0800 (PST) Received: from owen.delong.com (owen.delong.com [IPv6:2620:0:930::200:2]) by ietfa.amsl.com (Postfix) with ESMTP id 79E94C14CE53; Mon, 11 Dec 2023 11:11:55 -0800 (PST) Received: from smtpclient.apple (75-10-5-143.lightspeed.sntcca.sbcglobal.net [75.10.5.143]) (authenticated bits=0) by owen.delong.com (8.17.1/8.15.2) with ESMTPSA id 3BBJBsZY2683835 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Mon, 11 Dec 2023 19:11:55 GMT DKIM-Filter: OpenDKIM Filter v2.11.0 owen.delong.com 3BBJBsZY2683835 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=delong.com; s=mail; t=1702321915; bh=/20bthyLE3mUQy8ns14DeaIlrA5xmChvAOPco20/FCg=; h=From:Subject:Date:In-Reply-To:Cc:To:References:From; b=hbKvAMAWoa1pM7tg39uscGwYkS4ITURytPONdWLIi5t72xttQeO4Bw1gv22pFHCVt EUTqzfdm8+YeIP8bPA99QOzS3mNSykgKp8WYGzBKTPiHTmnWoRH/DpbT/XI0QCqWSG 83LCRCAtCTtY+dxYm+dz+9nQiHWJePp2PiJLFQaw= From: "owen@Delong.com" Message-Id: <09AAEFDB-C4A0-447B-844F-7DC9A099D7F1@delong.com> Content-Type: multipart/alternative; boundary="Apple-Mail=_4B185D95-AB47-4D0C-8266-47957D0B71F1" Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3774.200.91.1.1\)) Date: Mon, 11 Dec 2023 11:11:44 -0800 In-Reply-To: Cc: "owen@Delong.com" , list To: Daryll Swer References: <16354BB1-EB7A-4C80-921C-265A3A2FBA10@apnic.net> <968F3FB9-9DBC-4B87-A0F7-E9D91EB33492@massar.ch> <981F68FF-C61C-49E6-9E02-E9390142C263@delong.com> X-Mailer: Apple Mail (2.3774.200.91.1.1) X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.6.4 (owen.delong.com [192.159.10.2]); Mon, 11 Dec 2023 19:11:55 +0000 (UTC) Archived-At: Subject: Re: [v6ops] How was a /16 of IPv6 asigned from ARIN to non large network? X-BeenThere: v6ops@ietf.org X-Mailman-Version: 2.1.39 Precedence: list List-Id: v6ops discussion list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 11 Dec 2023 19:12:00 -0000 --Apple-Mail=_4B185D95-AB47-4D0C-8266-47957D0B71F1 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=utf-8 > > And again, back to IPv4 think. Why not /48 per customer site whether = enterprise, residential, or mobile? > > And that=E2=80=99s unfortunate for your customers and their = customers. >=20 > This isn't due to IPv4 thinking, yes a /48 per customer site for all = customers is awesome, I fully endorse this idea, in theory only. I would = love my residential ISP in my small home town to give me a /48 routed = from their PA as well. Well, HE will, FWIW. Mine wouldn=E2=80=99t either, so I went to ARIN instead. > But in reality, a /48 for everybody =3D massive RIR Fees for larger = allocations that simply, financially, is not justifiable for a lot of = network operators in a lot of economies, the financial numbers just = don't make sense. It really doesn=E2=80=99t. A /48 standalone is $250/year from ARIN. You get 16x that for 2x the fees at every higher level, so a /44 is = $500, a /40 is $1,000, 36 is $2,000, 32 is $4,000,=20 Let=E2=80=99s assume an average residential customer is US$30/month or = equivalent. That=E2=80=99s $360/year. Let=E2=80=99s assume that $359 of = that goes to circuit maintenance, backbone, and other costs leaving just = $1/year/customer to cover RIR fees. Further, let=E2=80=99s assume that your network structure doesn=E2=80=99t = allow you to do any better than 50% density in your /48 assignments = overall. That gives you 32,768 customers per /32. At that rate, you only need $0.125/customer/year to cover that /32. If = you have 16 times that many customers and need a /28, then you=E2=80=99ll = be paying $8,000/year to cover 524,288 customers =3D = $0.016/customer/year (rounded up from 0.0152587890625). APNIC=E2=80=99s fees are (IMHO) steep, but in APNIC, you=E2=80=99ve = chosen a governance model that allows the largest ISPs to dominate by = having a huge voting status where the large players (mostly government = run NIRs) get 64 votes vs. most smaller players only getting 1 vote. = This antidemocratic system creates many barriers to good customer = practices. Unfortunately, those in power are loathe to change the system, because = it keeps them in power. > Not all orgs have American-oriented financial power to justify paying = for larger RIR allocations. I know many ISPs with tens of thousands of = customers, they can't justify paying for large allocations to support = the /48 idea, many of them even struggle to afford two /32s let alone = anything larger. For example, in India, transport circuits+IP Transits = etc alone are super expensive, way too expensive for the tiny ARPU over = here to justify. And unfortunately, even though I at the very least = advocate for /56 minimum on wireline services, end-users simply don't = care, which disincentivize the ISP from investing any additional money = into RIR fees for larger blocks to grant /48s. For additional context, = many ISPs out there struggle to even purchase a router with decent FIB = capacity for full tables, some struggle to even deploy layer 3 PE = routers for the last-mile and still use some cheap Chinese OEM layer 2 = switches with STP etc. So you can see why /48 for everybody simply just = doesn't work, financially. Unfortunately, there=E2=80=99s little that can be done for unsustainable = business models where the ARPU is insufficient for the costs of = providing the service. Further, I think that the way APNIC has = structured its cost model is horrible. India does have an NIR, not sure = if that can be helpful for this or not. > Some other folks from the industry have shared this dilemma in the = past as well, including (but not limited to) AFRINIC-covered economies, = reference . Don=E2=80=99t get me started on the incredible corruption tax in = AFRINIC. Owen --Apple-Mail=_4B185D95-AB47-4D0C-8266-47957D0B71F1 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=utf-8
> And again, back to IPv4 = think. Why not /48 per customer site whether enterprise, residential, or = mobile?
> And that=E2=80=99s unfortunate for your customers and = their customers.

This isn't due to IPv4 thinking, = yes a /48 per customer site for all customers is awesome, I fully = endorse this idea, in theory only. I would love my residential ISP in my = small home town to give me a /48 routed from their PA as = well.

Well, HE = will, FWIW.
Mine wouldn=E2=80=99t either, so I went to ARIN = instead.

But in reality, a /48 for = everybody =3D massive RIR Fees for larger allocations that simply, = financially, is not justifiable for a lot of network operators in a lot = of economies, the financial numbers just don't make = sense.

It = really doesn=E2=80=99t.

A /48 standalone is = $250/year from ARIN.
You get 16x that for 2x the fees at every = higher level, so a /44 is $500, a /40 is $1,000, 36 is $2,000, 32 is = $4,000, 

Let=E2=80=99s assume an average = residential customer is US$30/month or equivalent. That=E2=80=99s = $360/year. Let=E2=80=99s assume that $359 of that goes to circuit = maintenance, backbone, and other costs leaving just $1/year/customer to = cover RIR fees.

Further, let=E2=80=99s assume = that your network structure doesn=E2=80=99t allow you to do any better = than 50%  density in your /48 assignments overall. That gives you = 32,768 customers per /32.

At that rate, you = only need $0.125/customer/year to cover that /32. If you have 16 times = that many customers and need a /28, then you=E2=80=99ll be paying = $8,000/year to cover 524,288 customers =3D $0.016/customer/year (rounded = up from 0.0152587890625).

APNIC=E2=80=99s fees = are (IMHO) steep, but in APNIC, you=E2=80=99ve chosen a governance model = that allows the largest ISPs to dominate by having a huge voting status = where the large players (mostly government run NIRs) get 64 votes vs. = most smaller players only getting 1 vote. This antidemocratic system = creates many barriers to good customer = practices.

Unfortunately, those in power are = loathe to change the system, because it keeps them in = power.

Not all orgs have = American-oriented financial power to justify paying for larger RIR = allocations. I know many ISPs with tens of thousands of customers, they = can't justify paying for large allocations to support the /48 idea, many = of them even struggle to afford two /32s let alone anything larger. For = example, in India, transport circuits+IP Transits etc alone are super = expensive, way too expensive for the tiny ARPU over here to justify. And = unfortunately, even though I at the very least advocate for /56 minimum = on wireline services, end-users simply don't care, which disincentivize = the ISP from investing any additional money into RIR fees for larger = blocks to grant /48s. For additional context, many ISPs out there = struggle to even purchase a router with decent FIB capacity for full = tables, some struggle to even deploy layer 3 PE routers for the = last-mile and still use some cheap Chinese OEM layer 2 switches with STP = etc. So you can see why /48 for everybody simply just doesn't work, = financially.

Unfo= rtunately, there=E2=80=99s little that can be done for unsustainable = business models where the ARPU is insufficient for the costs of = providing the service. Further, I think that the way APNIC has = structured its cost model is horrible. India does have an NIR, not sure = if that can be helpful for this or not.

Some = other folks from the industry have shared this dilemma in the past = as well, including (but not limited to) AFRINIC-covered economies, reference.

Don=E2=80=99t get me started on the = incredible corruption tax in = AFRINIC.

Owen


= = --Apple-Mail=_4B185D95-AB47-4D0C-8266-47957D0B71F1--