平 成 19 年 7 月 26 日 ネットワークの 中 立 性 に 関 する 懇 談 会 報 告 書 案 に 対 する 意 見 の 公 表 総 務 省 は ネットワークの 中 立 性 に 関 する 懇 談 会 ( 座 長 : 林 敏 彦 放 送 大 学 教 授 )において 取 りま とめた 報 告 書 案 について 平 成 19 年 6 月 22 日 から 同 年 7 月 23 日 までの 間 意 見 募 集 を 行 いました その 結 果 34 件 の 意 見 が 提 出 されましたので 公 表 します 1 経 緯 総 務 省 は 平 成 18 年 11 月 15 日 から ネットワークの 中 立 性 に 関 する 懇 談 会 を 開 催 し ネッ トワークのコスト 負 担 の 公 平 性 及 びネットワークの 利 用 の 公 平 性 を 軸 にネットワークの 中 立 性 につ いて 検 討 を 行 ってきたところです 本 懇 談 会 で 取 りまとめられた 報 告 書 案 について 平 成 19 年 6 月 22 日 から 同 年 7 月 23 日 までの 間 意 見 募 集 を 行 ったところ 34 件 の 意 見 が 提 出 されました 意 見 提 出 者 は 別 紙 のとおりです (なお 寄 せられた 意 見 については 準 備 が 整 い 次 第 総 務 省 ホームページ(http://www.soumu.go.jp)の 報 道 発 表 欄 及 び 電 子 政 府 の 総 合 窓 口 (http://e-gov.go.jp)の パブリックコメント 欄 に 掲 載 するとともに 連 絡 先 窓 口 において 配 布 します ) 2 今 後 の 予 定 皆 様 から 寄 せられた 意 見 を 踏 まえ 本 年 9 月 に 開 催 予 定 の 第 8 回 懇 談 会 において 報 告 書 を 取 りま とめる 予 定 です 関 係 報 道 資 料 ネットワークの 中 立 性 に 関 する 懇 談 会 報 告 書 案 の 公 表 及 び 本 案 に 対 する 意 見 の 募 集 ( 平 成 19 年 6 月 22 日 ) URL:http://www.soumu.go.jp/s-news/2007/070622_3.html ネットワークの 中 立 性 に 関 する 懇 談 会 の 開 催 ( 平 成 18 年 10 月 31 日 ) URL:http://www.soumu.go.jp/s-news/2006/061031_1.html 新 競 争 促 進 プログラム 2010 の 公 表 ( 平 成 18 年 9 月 19 日 ) URL:http://www.soumu.go.jp/s-news/2006/060919_4.html 本 懇 談 会 及 び 両 作 業 部 会 の 配 布 資 料 議 事 概 要 については 次 の URL を 御 参 照 ください URL:http://www.soumu.go.jp/joho_tsusin/policyreports/chousa/network_churitsu/index.html 連 絡 先 総 合 通 信 基 盤 局 電 気 通 信 事 業 部 料 金 サービス 課 大 矢 課 長 補 佐 竹 中 係 長 川 野 係 長 TEL:03-5253-5845 FAX:03-5253-5848
別 紙 意 見 提 出 者 一 覧 計 34 件 ( 受 付 順 敬 称 略 ) Ⅰ 電 気 通 信 事 業 者 等 19 件 受 付 意 見 提 出 日 ( 注 ) 意 見 提 出 者 代 表 者 氏 名 等 1 H19 年 7 月 20 日 北 海 道 総 合 通 信 網 株 式 会 社 取 締 役 社 長 高 野 幸 豊 2 H19 年 7 月 22 日 ロージナ 茶 会 3 H19 年 7 月 23 日 社 団 法 人 日 本 インターネット 会 長 渡 辺 武 経 プロバイダー 協 会 4 H19 年 7 月 23 日 KVH 株 式 会 社 代 表 取 締 役 社 長 鈴 木 みゆき 5 H19 年 7 月 23 日 ヤフー 株 式 会 社 代 表 取 締 役 井 上 雅 博 6 H19 年 7 月 23 日 株 式 会 社 エヌ ティ ティ ドコモ 代 表 取 締 役 社 長 中 村 維 夫 7 H19 年 7 月 23 日 ソフトバンクBB 株 式 会 社 代 表 取 締 役 社 長 孫 正 義 ソフトバンクテレコム 株 式 会 社 ソフトバンクモバイル 株 式 会 社 兼 CEO 8 H19 年 7 月 23 日 株 式 会 社 USEN 代 表 取 締 役 社 長 宇 野 康 秀 9 H19 年 7 月 23 日 日 本 電 信 電 話 株 式 会 社 代 表 取 締 役 社 長 三 浦 惺 10 H19 年 7 月 23 日 KDDI 株 式 会 社 代 表 取 締 役 社 長 小 野 寺 正 11 H19 年 7 月 23 日 富 士 通 株 式 会 社 代 表 取 締 役 社 長 黒 川 博 昭 12 H19 年 7 月 23 日 株 式 会 社 ウィルコム 代 表 取 締 役 社 長 喜 久 川 政 樹 13 H19 年 7 月 23 日 社 団 法 人 日 本 経 済 団 体 連 合 会 部 会 長 前 田 忠 昭 情 報 通 信 委 員 会 通 信 放 送 政 策 部 会 14 H19 年 7 月 23 日 社 団 法 人 テレコムサービス 協 会 政 策 委 員 会 15 H19 年 7 月 23 日 西 日 本 電 信 電 話 株 式 会 社 代 表 取 締 役 社 長 森 下 俊 三 16 H19 年 7 月 23 日 東 日 本 電 信 電 話 株 式 会 社 代 表 取 締 役 社 長 髙 部 豊 彦 17 H19 年 7 月 23 日 エヌ ティ ティ コミュニケーションズ 代 表 取 締 役 社 長 和 才 博 美 株 式 会 社 18 H19 年 7 月 23 日 イー アクセス 株 式 会 社 イー モバイル 株 式 会 社 代 表 取 締 役 社 長 代 表 取 締 役 社 長 安 井 敏 雄 エリック ガン 19 H19 年 7 月 23 日 モバイル コンテンツ フォーラム 座 長 東 邦 仁 虎 Ⅱ 個 人 受 付 意 見 提 出 日 ( 注 ) 1 H19 年 6 月 24 日 個 人 2 H19 年 6 月 24 日 個 人 14 件
3 H19 年 6 月 24 日 個 人 4 H19 年 6 月 24 日 個 人 5 H19 年 6 月 25 日 個 人 6 H19 年 6 月 25 日 個 人 7 H19 年 6 月 25 日 個 人 8 H19 年 7 月 2 日 個 人 9 H19 年 7 月 3 日 個 人 10 H19 年 7 月 12 日 個 人 11 H19 年 7 月 16 日 個 人 12 H19 年 7 月 23 日 個 人 13 H19 年 7 月 23 日 個 人 この 他 賛 成 する という 意 見 の 提 出 が1 件 ありました ( 参 考 ) 英 文 による 提 出 Ⅲ 英 文 受 付 意 見 提 出 日 ( 注 ) 意 見 提 出 者 1 H19 年 7 月 23 日 AT&T Inc. AT&T Japan Ltd. 1 件 ( 注 ) 意 見 提 出 日 は 総 務 省 に 提 出 された 日 ( 受 付 日 )を 記 載 しています
1/2
2/2
1
3 3 IP IP 5 3 2 3 4 5 1 2 3 2
4 5 24 IP IP IP 24 IT W 3
ISP 27 ISP 27 ISP ISP ISP FW 4
1. 2. 1 2 3 3. 4. 1. 2. 3. 1. 2. 3.( ) 3 P5 IP NGN NTT IP NGN FMC NTT 2
4 1. 2. 3. 4. 9 5 3
(2) (packet shaping) 1) (P22) (b) ISP ISP (P23) 2) (P24) ISP ISP ISP ISP ISP JAIPA ISP ISP ISP ISP ISP ISP ISP ISP 2
(P25) 4-1) (P27) P2 () ISP ISP ISP ISP audit ISP ISP ISP ISP ISP ISP ISP ISP ISP end-end 3
4-2) (p28) 4-3) (P29) ISP 14 USEN USEN ADSL ISP FTTH ISP ISP USEN 44 26 IP IP ISP IP 4
5 1. 2. P32 NTT
2-1) P34 (3) P37 (4) (P38) NNI SNI IP NTT IP IP NTT ISP IP ISP NGN 6
( ) IP ( ) 1) P41 4 1. (3) (P58) 2. (P59) NGN FMC ISP 7
ISP 8
1 2 20 ISP 25
3 25 QoS 36
4 IP 59
5
1
2
3
4
5
6
7
8
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
1. [1] 2. [1] [1] [1] 2.1. 1 N 1 N ()X B 1 B 2 DSL X NTT BB 4 [2] N 1 B, 2.2. 2 X N FTTH X N NTT 2 [2] OSU X N 1 2 3. 3.1. (1w ) 3 X 3 N 1 / N 2 / N 3 X A i B j S ij A 0 /A 1 /A 2 /A 3 P 0 /P 1 /P 2 /P 3 B 0 /B 1 Q 0 /Q 1 S ij 1
S S S S S 00 01 11 21 31 = P + Q, 0 1 0 2 3 1 0 = P + Q, 1 = P + Q, = P + Q 1, = P + Q. 1 (3.1) A i B j D ij a 1 c 0<4c<1 D D D D D 00 01 11 21 31 = a S = a S = a S = a S 00 01 11 = a S 21 31 + c( S + c( S + c( S 00 + c( S + c( S 01 00 00 00 + S + S + S + S 11 11 01 + S 01 01 + S + S + S + S 21 21 21 + S 11 11 + S + S + S + S + S 31 31 31 31 21 ), ), ), ), ). (3.2) D D D D D D A0 A1 A2 A3 B0 B1 = D = D = D 00 11 = D = D = D 21 31 00 01, + D,,, + D 01 11, + D 21 + D 31. (3.3) 2 A 0 /B 0 /B 1 A 1 / A 2 / A 3 N 1 / N 2 / N 3 X N 1 / N 2 / N 3 N1 / N2 / N3 Π Π Π Π X N1 N 2 N3 = P D 0 = P D 1 2 3 A0 = P D = P D A1, A2, A3 + Q. 0 D B0 + Q D 1 B1, (3.4) P 0 /Q 0 /Q 1 N 1 P 1 N 2 P 2 N 3 P 3 (1w ) X N1 / N2 / N3 D 00 D 01 D 11 D 21 D 31 D 1w X N1 / N2 / 3 (1w ) N3 P 0 /P 1 /P 2 /P 3 /Q 0 /Q 1 Π P X 0 Π = Q Π X P 0 X 0 2(1 7c) Q Π P 1 Π P Π Π X = Q = 2a 4(1 c) P 0 1 Π = P 0 2(1 c) Q N1 1 = 0, Π = P N 2 2 Π = P + 2cP + 2cP + 2cP 1 1 2 3 N 3 = a + 2cP0 2P1 + cp2 + cp3 (1 3c) Q0 + cq1 N1 = N2 2 N3 P 3 Π Q X 0 3 = 0. = a + 2cP0 + cp1 2P2 + cp3 (1 3c) Q0 + cq1 = 0, = a+ 2cP0 + cp1 + cp2 2P3 (1 3c) Q0 + cq1 = 0, = 4a 2(1 7c) P (1 3c) P (1 3c) P (1 3c) P 8(1 3c) Q + 8cQ = 0, Π X Q 1 0 1 0 = a 2(1 c) P0 + cp1 + cp2 + cp3 + 8cQ0 2Q1 a a P0 =, P 1 = P2 = P3 =, 2(3 2c) 3 2c a(1 + c) Q0 = Q1 =. (3 2c )( 1 4c) 1 2 0, = 0. (3.1) (3.2) 3 2
D D 00 11 = D = D 01 21 a( c+ 3) =, 2(3 2c) = D 31 a =. (3 2c) (1w ) 1 ' D w = D 00 + D 01 + D a(6 + c) =. (3 2c) 3.2. (2w ) 11 + D 21 + D X N 4 A i B j S ij A 1 A 2 P 1 P 2 B 1 B 2 Q 1 Q 2 S ij S 11 = P1 + Q1, S22 = P2 + Q2. (3.5) A i B j D ij a 1 c 0<c<1 D D 11 22 = a S = a S 11 22 + cs 22 + cs 11,. 31 (3.6) D D A1 A2 = D = D B1 B2 = D 11 = D, 22. (3.7) X A 1 /B 1 N A 2 /B 2 A/B X N Π Π X N = = P P 1 2 D D A1 A 2 + Q 1 + Q D 2 B1 D, B 2. (3.8) P 1 /Q 1 P 2 /Q 2 (2w ) 4 (2w ) X N D 11 D 22 D 2w X N P 1 / P 2 /Q 1 /Q 2 Π P X 1 Π X P 1 Π P N 2 = Π X = Q Π Q 1 Π = Q N 2 X 1 = Π P N 2 = Π Q N 2 = 0. = a 2( P1 + Q1 ) + c( P2 + Q2 ) = 0, = a + c( P1 + Q1 ) 2( P2 + Q2 ) = 0. S 11 a = P1 + Q1 =, S 2 c 22 a = P2 + Q2 =. 2 c (3.6) D 11 D 22 D 11 = a S a =. 2 c D 22 = a S a =. 2 c 11 22 + cs 22 + cs 11 (2w ) 3
2 ' D w 2a =. 2 c = D 11 + D 22 3.3. c 1 c 2 D 1w' a(6 + c1) = (3 2c ) D 1 1 2w' 2a 2 c a(6c1 c1c2 6c2 = (3 2c )(2 c 2 2 + 6). ) 0 < c 1 < 1/40 < c 2 < 1 D 2w <D 1w. 4. [1] 1 3 2 [1] pp.65-71 (2007). [2], (2006 4 (3 )) 19 6 18 (2007). 4
w D a c
S S S S S 00 01 11 21 31 = = = = = P P P 0 1 P P 0 2 3 + Q + Q + Q 1 + Q 0 1 + Q 1, 1,,, D D D D D D A0 A1 A2 A3 B0 B1 = D = D = D = D = D = D 00 11 21 31 00 01,,, + D, + D 01 11, + D 21 + D 31 D = a S + c( S + S + S + S ), D D D D D Π Π Π Π 00 01 11 21 31 X 1w' 00 a( c + 3) = a S = a S = a S = a S N1 N 2 N 3 2(3 2c) 1 3 00 01 11 2 21 31 = P D + c( S + c( S 00 + c( S + c( S A1, A2, A3 01 00 00 00 + S + S + S + S 11 11 01 01 01 D + S + S + S + S 1w' 11 21 21 21 11 11 + S + S + S + S 31 31 31 31 21 0 A0 0 B0 1 B1, = P D = P D = P D D 1w' 01 + Q D + Q D a( c + 3) a 2(3 2c) ( 3 2c) D ), ), ), ). 1w' 21 a D 1w' 31 a ( 3 2c) ( 3 2c) 1w' D a(6 + c) (3 2c)
S S 22 11 P1 + Q1, = = P + Q 2 2 D D A1 = DB 1 = D11, A2 = D B2 = D 22 D D 11 Π Π 22 X N = a S = a S = = 2 11 22 A2 + cs + cs 2 22 11 1 A1 + Q1D B1, P D P D + Q D B2 D 2w' 11 a 2 c D a 2w' 22 2 c 2w' D 2a 2 c
1 ' D w 2 ' D w a(6 + c1 ) = (3 2c1 ) 2a =, 2 c 2, 0 < 4 c 1 < 0 < c 2 < 1. 1. D = = a(6 + c (3 2c a(6c 1w' 1 1 1 ) ) 1 (3 2c D c c 1 2 2w' 2a 2 c 6c 2 2 )(2 c 2 + 6) ) > 0 1w' 2w' D D 2w' 1w' D < D c 2 c 1
1. packet shaping
IP 30 2
1
2
3
4
5
6
7
1
2
3
4
5
6
7
1
2
1
2
3
4
5
1
1 ) NTT 2
NTT IP 3
4 NTT MNO ID FMC 4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1 ()
2
3
4
5
6
7
8
9
10
11
12
13
1
2
3
4
5
6
7
2
3
4
5
6
7
8
P2P YouT ube http ISP winny WinMX ISP 2
ISP 100M ISP ISP NTT CM ISP Q2 3
IT ISP 4
5
別 紙 2 章 第 1 章 ネットワークの 中 立 性 に 関 する 基 本 的 視 点 第 2 章 ネットワークのコスト 負 担 の 公 平 性 第 3 章 ネットワークの 利 用 の 公 平 性 1. 検 討 の 目 的 2.ブロードバンド 市 場 における 環 境 変 化 3. 検 討 に 際 しての 基 本 的 視 点 4. 諸 外 国 におけるネットワーク の 中 立 性 を 巡 る 議 論 1.ネットワークのコスト 負 担 の 公 平 性 を 確 保 するための 基 本 的 視 点 2.ネットワーク 上 のトラフィックの 現 状 3.ネットワークのコスト 負 担 の 公 平 性 に 係 る 検 討 課 題 1.ネットワークの 利 用 の 公 平 性 を 確 保 するための 基 本 的 視 点 2. 次 世 代 ネットワークに 係 る 接 続 ルールの 在 り 方 具 体 的 内 容 別 紙 1ページのコメント1 参 照 別 紙 1ページのコメント2 参 照 別 紙 1ページのコメント3 参 照 別 紙 1ページのコメント4 参 照 第 4 章 ネットワークの 中 立 性 に 関 するその 他 の 検 討 課 題 3.ドミナント 規 制 ( 指 定 電 気 通 信 設 備 制 度 )の 在 り 方 1.ネットワークの 中 立 性 に 関 す るその 他 の 検 討 課 題 2. 利 用 者 保 護 策 の 検 討 3. 端 末 政 策 の 見 直 し 4. 国 際 的 な 制 度 の 整 合 性 確 保 への 積 極 的 貢 献 第 5 章 望 ましい 政 策 展 開 の 方 向 性 補 論 ドミナント 規 制 の 見 直 しと 競 争 評 価 の 活 用 の 在 り 方 別 紙 1ページのコメント5 参 照 < 記 載 要 領 > 具 体 的 内 容 欄 に 御 意 見 の 具 体 的 内 容 を 御 記 入 ください
3 コメント1 データの 記 載 方 法 に, 曖 昧 さがある. 特 に,P.17 の P2P トラフィックの 占 有 率 は,2003 年 6 月 時 点 と 明 記 すべきである. 重 要 なのは,インターネットはキラーアプリが 出 るとトラフィックパターンが 劇 的 に 変 わる ことであり,2007 年 時 点 では,ひょっとすると GAO や U Tube からのビデオのダウンロードの 方 が 支 配 的 かも しれない. かもしれない と 言 わざるを 得 ないところに 隘 路 があり,このような 問 題 の 議 論 は,もっと 正 確 な 数 値 データに 裏 づけされてできるようにならなければならない. コメント2 ベストエフォート サービスでのトラフィックの 弁 別 は, 好 ましいものではなく, 基 本 はやらないこと と 考 える. 安 易 なトラフィック 制 御 は, 新 しいサービスの 芽 を 摘 むことにもなりかねない.また,インター ネットにおけるトラヒックのアンバランスは, 黎 明 期 から 続 いており,P2P に 限 ったことではない. 実 際 初 期 のインターネットでもメールと NetNews を 比 べたら,NetNews の 利 用 者 数 はメールの 利 用 者 数 に 比 べれば 桁 違 いに 少 なかったが,トラフィックの 大 勢 を 占 めていた.また, 使 っている 技 術 は 異 なるが, 不 特 定 多 数 の 興 味 を 持 った 人 へのコンテンツ 配 信 ということでは, 今 日 の P2P と 根 源 的 思 想 は 同 じである.この 種 の 技 術 は 時 と 形 を 変 えていつでも 生 まれてくるものと 覚 悟 しなければならない. コメント3 Everything over IP 時 代 を 迎 えた 今,これまでの 枠 でネットワークの 中 立 性 を 論 ずるのは 意 味 がない. 直 接 的 通 信 設 備 を 持 つキャリアとそうではない ISP に 分 けて 考 えるべきである.キャリアの 場 合, 従 来 の 専 用 線 等 を 使 って 行 ってきたサービスをも IP インフラ 上 に 移 行 しなければならないが,インターネット 用 と 設 備 を 分 離 するのは,OPEX も CAPEX も 上 昇 するだけである.サービス 価 格 の 上 昇 につながり, 最 終 的 な 顧 客 の 満 足 は 得 られない. 従 って, 設 備 共 用 はビジネスの 自 然 な 帰 結 である.また,NGN はその 走 りと 考 える. 問 題 は,そのような 高 品 質 サービスが 導 入 されたことにより,ベストエフォートサービスがどれだけ 割 を 食 うことになるのかである. 現 在 の IP 技 術 で QoS を 制 御 する 場 合,スケーラビリティの 観 点 からパケットに 優 先 順 位 付 けすることくらいしかできない. 過 去 30 年 以 上 に 渡 って,パケット 通 信 の 品 質 向 上 が 叫 ばれ, 研 究 されてきた 結 果 である.この 程 度 の 制 御 技 術 では, 優 先 パケットは 安 全 を 見 て 平 均 40-50%の 回 線 占 有 率 に 抑 えておかないと, 満 足 のいく 通 信 品 質 は 保 てない. 換 言 すれば, 優 先 パケットの 品 質 は,それと 等 量 以 上 のベストエフォートのパケットが 存 在 することによって, 保 証 されるとも 言 える. 優 先 パケットが 導 入 され たとしても, 大 部 分 ( 半 分 以 上 )のトラフィックがベストエフォートならば, 従 来 のインターネットのフレ ームワークは 崩 れることはない.つまり,NGN は, 同 一 設 備 上 でサービスされるベストエフォートのインター ネットと 共 存 可 能 である. コメント4 問 題 は,ドミナント 事 業 者 が, 優 先 サービスを 自 社 サービスやグループ 会 社 の ISP へ 選 択 的 に 提 供 するよ うなことが 避 けられること, 中 小 の 優 先 サービス 提 供 業 者 との 相 互 接 続 性 を 担 保 すること,これがネットワ ークの 中 立 性 を 論 ずる 際 の 核 であると 思 う.たとえば,Skype 常 用 者 なら 誰 でも 知 っている 事 実 であるが, Skype は 音 質 そのものはいいが, 音 がしばしば 飛 ぶため, 長 時 間 の 駄 弁 りには 良 いが, 電 話 会 議 には 使 えない. 従 って, 取 締 役 会 の 記 録 にも 耐 えるようなサービスにするには, 優 先 サービスの 導 入 は 避 けられない.これ が 同 一 通 信 設 備 上 の 1 社 にしかできないようでは, 公 正 な 競 争 に 持 っていけないであろう. コメント5 今 回 の 報 告 で 分 かったことは,(1) 議 論 に 足 るだけのトラフィック 情 報 を 入 手 できないこと,(2)ISP はすで に 種 々のパケット フィルタリングを 実 施 しだしていることである. 前 者 に 関 しては, 新 たな 通 信 データ 公 開 原 則 のようなものが 必 要 であろう.また, 後 者 に 関 しては, 通 信 の 秘 密 の 観 点 からその 種 の 運 用 がどこま で 許 されるのか 否 か 国 民 のコンセンサスが 必 要 だと 考 える.
AT&T Japan Comments on the Draft Report on Net Neutrality AT&T Japan Ltd. ( AT&T Japan ) is pleased to submit these comments on the Ministry of Internal Affairs and Communications ( MIC ) draft report on net neutrality published on June 22, 2007 (the Draft Report ). AT&T Japan is owned by AT&T Inc. ( AT&T ) and provides international business services in Japan. AT&T, through its affiliates, operates one of the world s most advanced global backbone networks, provides services to virtually every country and territory in the world, and is a leading provider of international business and consumer communications services on the U.S.-Japan route. AT&T Japan appreciates this opportunity to provide further comment to the MIC Study Group on this important issue. I. INTRODUCTION AND SUMMARY AT&T Japan endorses the core conclusion of the Draft Report that pricing and traffic arrangements among Internet providers should continue to evolve and be governed by commercial negotiations, rather than ex ante regulatory mandates. We look forward to assisting the MIC Study Group with its further analysis of the highly dynamic and fastevolving Internet marketplace. As AT&T Japan has previously described in this proceeding, there is no market failure that could warrant regulatory intervention in the Internet, and the downside risk of such intervention would be enormous. As in other contexts where ex ante regulation could produce severe unintended consequences, market forces should be allowed to continue doing what they do best build consumer value and any intervention should be limited to ex post enforcement remedies if and when a market failure arises. Other policymakers are reaching similar conclusions. A June 2007 Staff Report, unanimously approved by the bipartisan U.S. Federal Trade Commission (FTC), found that, by all indications, the broadband Internet access industry is young and dynamic and is moving in the direction of more, not less, competition, including fast growth [and] declining prices for higher-quality services. 1 The FTC further explained that we are unaware of any significant market failure or demonstrated consumer harm from 1 FTC Staff Report, Broadband Connectivity Competition Policy 10-11 (June 27, 2007).
2 conduct by broadband providers. 2 The FTC warned that [p]olicy makers should be wary of enacting regulation solely to prevent prospective harm to consumer welfare, both because there is no demonstrated need for such regulation and because [i]ndustrywide regulatory schemes particularly those imposing general, one-size-fits-all restraints on business conduct may well have adverse effects on consumer welfare. 3 The FTC s carefully researched opposition to preemptive net neutrality regulation closely resembles the analysis announced three months ago by the Organisation for Economic Co-operation and Development (OECD), which found that [t]here is little evidence of anti-competitive conduct to date, and it seems premature for governments to become involved at the level of network-to-network traffic exchange and demand neutral packet treatment for content providers. 4 Indeed, the inability of supporters of regulatory intervention to demonstrate any evidence of an actual market failure confirms what the rhetoric on this issue cannot obscure: net neutrality is a solution in search of a problem. Moreover, regulatory intervention when there is not even a theoretical basis for concern that broadband providers will engage in anticompetitive conduct with traffic and pricing arrangements, given the dynamic and interdependent nature of the Internet marketplace, would be highly damaging. Net neutrality regulation would reduce competition on the Internet, and compromise its efficiency, by inhibiting operators of broadband access and backbone networks from competing with content delivery networks and others for the provision of performance-enhancing services to applications and content providers. Such regulation also would harm consumers in several independent ways. By forcing broadband networks to commoditize their services, it would both reduce consumer choice and undermine the incentives of broadband providers to continue investing billions of dollars in next-generation infrastructure. That investment-deterring effect would fall especially hard on the underserved communities and would undermine efforts to bridge the digital divide. The net neutrality agenda would force broadband 2 Id. at 11. 3 Id. (emphasis added). 4 OECD Report, Internet Traffic Prioritisation: An overview 5 (Apr. 6, 2007) (http://www.oecd.org/dataoecd/43/63/38405781.pdf).
3 providers to recover the network costs of accommodating bandwidth-intensive applications from subscribers alone rather than from the providers of those applications. It would thus raise broadband rates, depress broadband subscribership, and introduce debilitating inefficiencies into the Internet marketplace. Only if broadband providers retain flexibility to negotiate quality-of-service terms with the providers of these bandwidth-intensive applications will it be possible to achieve what should be the two overriding goals for the 21st century Internet: affordable consumer prices for basic broadband connectivity and the proliferation of high-quality real-time applications over the Internet platform. The Draft Report thus properly recognizes the importance of allowing pricing models to develop through market negotiations rather than through regulatory mandate. AT&T Japan does not support the suggestion by the Draft Report, however, that some regulatory guidelines may be appropriate to address packet shaping. As described in more detail below, IP was always intended to be a flexible protocol that could be adapted to meet the specific needs of the applications running over it, and packet differentiation has become a routine and increasingly essential aspect of network management to maintain the cost-reducing efficiencies of packet switched IP networks while assuring the delivery of high quality multimedia services to consumers. Even Google concedes that broadband networks should be free to prioritiz[e] packets associated with latency-sensitive applications 5 and that the Internet today is not an absolutely neutral place because its constituent networks (presumably including Google s) can and do distinguish routinely between various forms of traffic. 6 As described below, prohibiting all packet differentiation would have massive adverse consequencies for the global Internet, and efforts to draw a regulatory line between diffferentiation and discrimination are likely to be both unsuccessful and equally harmful. Thus, for the same reasons that MIC does not intend to regulate commercial arrangements for Internet traffic, it would not be appropriate or constructive to establish regulations governing packet shaping. 5 at 22. 6 Google Comments, FCC WC Docket No. 07-52, Broadband Industry Practices, filed Jun. 15, 2007, Id. at 4 n.6.
4 In particular, the nondiscrimination rules favored by some net neutrality advocates would chill the free-wheeling experimentation at the heart of the Internet s success and would embroil the industry in years of indeterminate litigation about the reasonableness of highly technical network-management decisions made in a rapidly evolving business environment. Further, no regulator could rationally impose nondiscrimination rules only on broadband Internet access providers without also applying those rules to other information service providers that exert greater structural influences on the Internet s development. For example, regulators would need to impose such rules on content delivery networks, which specialize in ensuring that the Internet will not treat all applications and content alike, and on Google and other providers of Internet advertising and search services. Indeed, more than any access or backbone network, Google affects where end users will go on-line and which Internet sites will succeed and which will fail, and which voices will be heard or not. AT&T Japan emphasizes that it does not advocate that any Internet-based company be subject to nondiscrimination rules: the Internet has succeeded precisely because governments have kept information services free from regulation. In sum, regulators should not take this road at all: instead they should maintain the free play of market forces in the Internet ecosystem. The following sections describe these concerns in more detail. II. MANAGING BANDWIDTH SCARCITY A. The Explosive Growth of Bandwidth-Intensive Internet Applications. Until recently, the most popular Internet activities consisted of surfing mostly text-oriented webpages, conducting basic e-commerce, exchanging e-mails, and downloading document files. Although none of these activities consumed enormous bandwidth by today s standards, networks still had to make substantial capacity upgrades over time in order to accommodate increasing traffic volumes. Access and backbone networks were built to accommodate those consumer needs quickly and efficiently, and they succeeded. Over the past few years, however, bandwidth-intensive applications such as streaming video have exploded in popularity. These applications have dramatically increased the volume of Internet traffic both downstream and upstream
5 and have begun placing unprecedented demands on the capacity of the Internet s access and backbone networks. YouTube graphically illustrates this phenomenon. That company, which Google recently acquired for $1.65 billion, did not even exist in January 2005. One year after its inception, in February 2006, it had begun purchasing backbone transit services for 20 Gbps of video traffic equivalent to about two million simultaneous emails with a growth rate of 20% compounded monthly. 7 Today, users download more than 100 million video clips from YouTube per day, 8 and the trajectory of that service s bandwidth consumption is staggering: Each year the original content on the world s radio, cable and broadcast television channels adds up to about 75 petabytes of data or, 10 to the 15th power. If current estimates are correct, the two-year-old YouTube streams that much data in about three months. But a shift to highdefinition video clips by YouTube users would flood the Internet with enough data to more than double the traffic of the entire cybersphere. 9 YouTube is just the vanguard of the streaming video business. More generally, Yankee Group projects that the number of video streams viewed will double over the next five years (from 86 billion in 2006 to 166 billion in 2011) and that the average length of each video stream will increase more than five-fold during the same period (from 4.87 minutes per stream to 25.09 minutes). 10 Other types of Internet-based video technologies and business models are rapidly taking root as well. For example, Joost created in January 2007 by the founders of Skype delivers peer-to-peer video programming under contractual arrangements with major U.S. programmers, including CBS, Viacom, Time Warner, Sony, and the National Hockey League. 11 Netflix plans to shift its massive distribution of feature-length movies 7 William B. Norton, Video Internet: The Next Wave of Massive Disruption to the U.S. Peering Ecosystem, v1.3, at 2 (Equinix 2007) ( Video Internet ). 8 Rob Hof, YouTube: 100 Million Videos a Day, BusinessWeek, July 14, 2006 (http://www.businessweek.com/the_thread/techbeat/archives/2006/07/youtube_100_mil.html). 9 Bret Swanson, The Coming Exaflood, Wall St. J., Jan. 20, 2007, at A11. 10 Yankee Group, 2006 Internet Video Forecast: Broadband Emerges as an Alternative Channel for Video Distribution 6-7 (Dec. 2006). 11 Alex Pham, Joost Strikes TV Deals with Sony, Time Warner, NHL and Hasbro, L.A. Times, May 2, 2007, at 3.
6 from the postal service to the Internet, 12 and Microsoft has announced that Xbox 360 users will soon be able to download movies directly to their gaming consoles over the Internet. 13 Overall, industry experts expect that video traffic will represent at least 80 percent of all Internet traffic by 2010. 14 Video is only one of several bandwidth-intensive applications that are placing new demands on the Internet s access and backbone networks. Others include such music downloading services as itunes, which have supplanted compact discs as the primary means of music distribution; on-line printing and photo-sharing services such as Kodak, Snapfish, Shutterfly, and Photobucket; and the enormously popular class of massively multiplayer online role-playing games, such as Sony s EverQuest and Blizzard Entertainment s World of Warcraft. These MMORPGs are graphics-intensive, three-dimensional on-line video games played simultaneously by thousands of end users dispersed throughout the Internet. Because the end users are playing the same games with one another in real time, these applications require an unusually high quality of service from the networks over which they are provided. 15 P2P technologies such as KaZaA and BitTorrent, which now account for a large percentage of all traffic on the publicly accessible Internet, 16 likewise present intense bandwidth challenges. These technologies disassemble content into small files and widely distribute those files to different end-user computers for storage and subsequent retrieval by other end users. 17 The result is the functional equivalent of a massively distributed server network. By converting end user devices into content caches for other end users, P2P technology offers a cheap distribution option for content providers, but imposes enormous upstream traffic burdens on Internet access networks, particularly with the rise of shared video. Peer-2-Peer file sharing users around the world have shifted from sharing 4MB music files to sharing 700MB movies on the Internet. These 12 See Press Release, Netflix Offers Subscribers the Option of Instantly Watching Movies on Their PCs, Jan. 16, 2007 (http://www.netflix.com/mediacenter?id=5384). 13 See Robert Levine, Microsoft to offer Xbox movie downloads, New York Times, Nov. 7, 2006 (http://www.iht.com/articles/2006/11/07/business/xbox.php). 14 Norton, Video Internet, supra, at 2. 15 See Robert Litan & Hal Singer, Unintended Consequences of Net Neutrality Regulation, (forthcoming J. Telecomm. & High Tech. L. (2007)). 16 David Vorhaus, Confronting the Albatross of P2P, Yankee Group (May 31, 2007). 17 See, e.g., Detlef Schoder, Kai Fischbach, & Christian Schmitt, Core Concepts in Peer-to-Peer Networking (2005) (http://www.idea-group.com/downloads/excerpts/subramanian01.pdf).
7 applications relentlessly consume all of the end user s available Internet bandwidth attempting to download chunks of the files from any sources on-line at the time.... [T]he access networks are filling up 24/7, resulting in significant costs to the accessheavy ISPs (Cable Companies and DSL Providers in particular). 18 The network-management challenges posed by bandwidth-intensive applications arise, moreover, not just from an increase in the total volume of Internet traffic, but also from the escalating magnitude of unpredictable spikes in Internet traffic. Like conventional telephone networks, IP networks are sized to handle demand during periods of peak usage. The closer that peak usage is to average usage, the more efficient the network s cost structure will be, and the more predictably the network operator can recover those costs from the users of its network. According to some estimates, however, video applications roughly double the peak-to-mean ratio of traffic on IP networks. 19 The reason relates in part to the viral (self-intensifying) nature of popular video files. A video clip accompanying a breaking news or entertainment story often triggers unexpected network congestion by generating massive Flash Crowd effects, and Viral Amplifiers (sites that do not host but rather highlight the most popular videos) amplify any viral properties a video may have. 20 Similar demand spikes can appear after the release of popular new products. For example, data traffic associated with Xbox Live surged 80% during the week after Microsoft released six new games for that platform in November 2006, and North American broadband service providers saw an approximate 140% peak increase attributed to itunes bandwidth usage just after Apple released itunes 7, enabling users to download full feature-length films onto their ipods. 21 Predictably, traffic associated with these same products surged again starting on Christmas day. 22 With less warning, Internet traffic would also spike in the wake of a pandemic, terrorist attack, or other emergency that suddenly causes millions of people to 18 See, e.g., William B. Norton, The Evolution of the U.S. Internet Peering Ecosystem, at 8 (2003) (http://www.equinix.com/pdf/ whitepapers/peeringecosystem.pdf). 19 See Norton, Video Internet, supra, at 3. 20 Id. at 1. 21 Sandvine Press Release, Sandvine Identifies Surge in Xbox Live and itunes Traffic (Dec. 4, 2006) (http://www.sandvine.com/news/pr_detail.asp?id=111). 22 Sandvine Press Release, Sandvine Observes Sustained Surge of Xbox Live and itunes Traffic (Feb. 5, 2007) (http://www.sandvine.com/news/pr_detail.asp?id=117).
8 turn to the Internet for information and, subsequently, to become telecommuters. For example, industry executives warn that, in the event of a bird flu outbreak, the demand for communication will soar, and the Internet s ability to handle that surge is definitely the most vulnerable part of the equation. 23 On top of the escalating volume of Internet traffic and the increasingly unpredictable spikes in that volume, an increasing number of applications rely on certain protocols that aggressively consume available bandwidth at the expense of other applications using different protocols. For example, the Internet is often described as using the TCP-IP protocol suite, with IP at Layer 3 (the network layer) and TCP at Layer 4 (the transport layer). TCP is considered a polite transport protocol because it can sense congestion and throttles back transmission rates until after the congestion lifts. But not all Internet applications use TCP. Some use the User Datagram Protocol ( UDP ) at Layer 4. UDP omits the error-correction functions of TCP and, unlike TCP, does not throttle back in the face of network congestion. 24 In the words of one network engineer, UDP is an inherently brutish, congestion ignorant protocol. It blindly blasts data at the rate desired with no built-in notion of loss or bandwidth hijacking. Effective measures must be added in order to send UDP datagrams in a congestion sensitive environment. 25 Moreover, even applications that use TCP can and do aggressively consume disproportionate amounts of subscriber bandwidth simply by opening up multiple streams (or torrents, as featured in the BitTorrent P2P technology) to seize capacity for themselves. 26 23 Patrick Thibodeau, Flu pandemic could choke Internet, requiring usage restrictions, ComputerWorld, Feb. 12, 2007 (http://www.computerworld.com/action/article.do? command=viewarticlebasic&articleid=9011125) (quoting Renate Noone, vice president of professional services at SunGard s Availability Services unit, and Bernard O Neill, vice president and chief network officer at Prudential Financial Inc.). 24 UDP applications thus send out data as fast as [they] can, even when they encounter congestion, while [conventional] TCP-friendly applications deliberately send fewer and fewer packets and may thus end up starved of network resources. Jon M. Peha, The Benefits and Risks of Mandating Network Neutrality, and the Quest for a Balanced Policy, at 7 (2006) ( Benefits and Risks ) (http://web.si.umich.edu/tprc/papers/2006/574/ Peha_balanced_net_neutrality_policy.pdf). Nonetheless, when properly managed, UDP s attributes can be beneficial for a range of purposes, including Domain Name System (DNS) queries. 25 William Strathearn, A Fair and Balanced Method of Controlling UDP Congestion (2003) (http://strathearns.org/wds/papers/udpcongestiioncontrol.pdf). 26 See, e.g., Bob Briscoe, Flow rate fairness: Dismantling a Religion, 37 Computer Commc n Rev. 63 (2007) (http://portal.acm.org/citation.cfm?doid=1232919.1232926).
9 Absent effective network-management measures, the bandwidth-hogging nature of UDP and some applications can create a classic tragedy-of-the-commons dynamic, in which each network user has every incentive to hoard shared bandwidth for itself at the expense of others. To avoid this result, each network provider the party ultimately responsible for managing shared network resources needs to ensure that sufficient network resources are available for all of its customers on an equitable basis and that no single customer (or group of customers) diminishes the value of the network for others. B. Tracking the Exaflood. As the preceding discussion makes clear, broadband providers are constantly challenged to keep up with the accelerating demands the new bandwidth-intensive applications are placing on their networks. Earlier this year, Deloitte Touche Tohmatsu reported a growing consensus among engineering experts that the upsurge in Internet traffic due to new bandwidth-intensive applications will likely require considerable investment in new infrastructure. 27 Deloitte noted the increase in the total number of Internet users, which breached the one billion mark in 2006, and the sudden popularity of video applications, including peer-to-peer video, video chat, video downloads, user-generated video content and IPTV. 28 Although Deloitte avoided taking sides in the net neutrality debate, it concluded that, [c]learly, something has to change in the economics of Internet access, such that network operators and ISPs can continue to invest in new infrastructure and maintain service quality, and consumers can continue to enjoy the Internet as they know it today. 29 There is nothing speculative about Deloitte s concern. Other parties, from all corners of the industry, have expressed exactly the same concern. Vincent Dureau, Google s head of TV technology, warned: The web infrastructure and even Google s doesn t scale. It s not going to offer the quality of service that consumers 27 Deloitte Touche Tohmatsu, Telecommunications Predictions: TMT Trends 2007, at 6 (2007) (http://www.deloitte.com/dtt/cda/doc/content/dtt_telecompredictions011107.pdf). 28 Id. 29 Id. at 7.
10 expect. 30 Former U.S. Assistant Secretaries of Commerce Larry Irving and Bruce Mehlman now co-chairmen of the Internet Innovation Alliance observed: [A]s new content proliferates, today s high-speed connection could be tomorrow's traffic jam. The strain on broadband capabilities and the looming data deluge is often called the Internet exaflood. Exaflood stems from the term exabyte, or 1.074 billion gigabytes. Two exabytes equal the total volume of information generated in 1999. The Internet currently handles one exabyte of data every hour. This mushrooming amalgamation of data is pushing the Internet to its limits.... The Internet infrastructure must be robust enough to handle all of the new data; this is often a challenge because the Internet is really thousands of privately owned, individual networks stitched together. It requires constant investment so that it will continue to grow and run smoothly. 31 And William Norton of Equinix, a leading provider of data centers and Internet exchange services, predicts a new wave of disruption that potentially dwarfs currently peered Internet traffic, causing a serious supply side problem in the Internet today. 32 C. Engineering Solutions to the Exaflood. The Internet can rise to the exaflood challenge, and, when it does, it will radically enhance the Internet experience. For the Internet to get from here to there, however, it must continue evolving in three basic ways. First, more IP networks need to be built in order to handle the surging data flow. Second, the data-carrying capacity on existing IP networks needs to be bigger. Third, each IP network also needs to be smarter. Edge providers like Google, Akamai, and Limelight have begun responding to the call for more IP networks. They have built the functional equivalent of new global backbones to ensure exceedingly efficient connectivity between the providers and end users of Internet applications and content. At the same time, policymakers have properly stressed the need for more access networks and capabilities to connect more consumers to 30 Bobbie Johnson, Rise of video downloads threatens gridlock on net, The Guardian, Feb. 10, 2007 (http://www.guardian.co.uk/uk_news/story/0,,2010031,00.html) (quoting Vincent Dureau, Google s head of TV Technology). 31 Bruce Mehlman & Larry Irving, Bring on the Exaflood!, Wash. Post, May 24, 2007, at A31 (http://www.washingtonpost.com/wp-dyn/content/article/2007/05/23/ AR2007052301418.html); see also Internet Innovation Alliance, Leading Internet Alliance Says Web s Capacity Is Nearing Its Limit, Apr. 18, 2007 (http://www.internetinnovation.org/iia/ page/show/news_clips_041807b). One exabyte equals 1024 petabytes, or 1,048,576 terabytes, or 1,073,741,824 gigabytes. [F]ive exabytes are equal to all the words ever spoken by human beings. Id. By comparison, [t]he information added annually to the digital universe between 2006 and 2010 is expected to increase more than six fold to 988 exabytes. Id. 32 Norton, Video Internet, supra, at 1, 15.
11 the Internet and to handle the accelerating traffic flows in the last mile. As we discuss below, the best route to that outcome is to promote a stable deregulatory environment that encourages such capital-intensive ventures. That is also the best way to encourage existing network providers to invest in making their pipes bigger. In the favorable U.S. deregulatory climate, those providers have invested many billions of dollars in facilities upgrades. For example, AT&T is investing $6.5 billion in its IP-based Project Lightspeed network, which extends fiberoptic cable deep into individual neighborhoods and will give tens of millions of Americans, among many other things, a much-needed alternative to the cable incumbents for the delivery of multi-channel video programming. 33 AT&T is also investing substantial sums to quadruple the capacity of its global backbone network by upgrading to OC-768 pipes with speeds of 40 gigabytes per second. 34 Verizon is investing billions of its own to deploy its FiOS broadband and video distribution network, which extends fiber-optic cables to individual homes and businesses. 35 U.S. cable providers, too, have reportedly spent more than $110 billion over the last decade to build out a 21 st century platform for advanced services. 36 At the same time, Sprint and Clearwire are investing heavily in WiMAX technology, and they are aggressively deploying nationwide broadband networks in the United States that will each be capable of serving more than 100 million customers. 37 Private companies have made these investments without any guarantee of costrecovery. Indeed, competition and the unpredictability of consumer demand make these investments however essential to the Internet s future exceptionally risky for the companies that underwrite them. Fearful that the telecom crash of the early 2000s will repeat itself, many on Wall Street respond with great skepticism to the increased capital spending needed for these capacity upgrades. For example, Comcast shares fell more 33 See Dionne Searcey & Peter Grant, AT&T Says Costs Rise for TV System s Launch, Wall St. J., May 8, 2007, at B4. 34 See Craig Matsumoto, AT&T Readies 40-Gig Backbone, Light Reading (June 6, 2006) (http://www.lightreading.com/document.asp?doc_id=96564&site=globalcomm). 35 See Verizon press release, Sept. 26, 2006 (http://investor.verizon.com/news/ view.aspx?newsid=773). 36 NCTA Comments, FCC WC Docket No. 07-45, at 2-3 (May 16, 2007). 37 See Sprint Comments, FCC WC Docket No. 07-45, at 8 (May 16, 2007); Clearwire Comments, FCC WC Docket No. 07-45, at 3, 5 (May 16, 2007).
12 than 3 percent on Feb. 1 when the company gave a higher than expected outlook of $5.7 billion on capital spend for 2007. 38 As a Sanford Bernstein analyst explained, Comcast can talk till it s blue in the face about what the marginal return on investor capital is, but investors are so nervous about capital spending that they re almost hoping for slower growth[.] 39 Likewise, as a Lehman Brothers analyst observes, [t]he market is very skeptical of [Verizon s] FiOS spending as well. 40 Wharton professor Kevin Werbach adds: Verizon is making a smart long-term bet on fiber to the home, but it s a risky bet. The capital costs are substantial and the payback scenarios are uncertain[.] 41 While these risky, capital-intensive capacity upgrades are necessary steps that any network provider must take to keep pace with escalating bandwidth demands, they are not sufficient steps. Each provider must also make its network smarter. Below we summarize why IP networks must be smart enough to treat packets differently, and then we address how they do so. 1. Why providers manage their IP networks. Internet-related service providers must carefully manage their networks in order to deliver reliable, high-quality services to their customers. That task is challenging precisely because the Internet Protocol has made the dream of convergence a reality. In the past, voice, video, and data services were, for the most part, provided over separate, single-purpose networks. IP changes all of that. Now, any type of electronic communications can be offered over a unified IP platform. This convergence of multiple services onto a single platform carries many advantages beyond the obvious economies of scale and scope derived from building one network rather than several. It also allows for the integration of voice, video, and text into feature-rich multimedia applications. And it facilitates greater competition among service providers. For example, cable and telephone companies, once siloed from mutual competition because of their single- 38 Yinka Adegoke (Reuters), Comcast sees cable rev rising, capex steady, May 1, 2007 (http://today.reuters.com/news/articleinvesting.aspx?type=companynews &storyid=220959+01-may-2007+rtrs&wtmodloc=invart-l2-companynews-3). 39 Id. 40 Ed Gubbins, Lehman: Verizon may want to rethink FTTP, Telephony Online, Jan. 4, 2006 (http://telephonyonline.com/fttp/news/lehman_verizon_fttp_010406/). 41 Knowledge@Wharton, Verizon s High-Speed Network: If They Build It, Will You Come? (Mar. 21, 2007) (http://knowledge.wharton.upenn.edu/article.cfm?articleid=1689).
13 purpose networks, now compete fiercely to offer the triple play of voice, video, and Internet access services. But the benefits of convergence come with a potential cost. Because related packets within an Internet data session may follow different routes to their destination, encounter congestion along the way, and arrive out of sequence, it may take a few moments for an end user s computer (or other device) to buffer, reorganize, and assure a consistent delivery rate for the packets received for given applications. The complexity of that process increases with delays in packet transmission ( latency ), inconsistent packet delivery intervals ( jitter ), and packet loss. The more time that packets spend en route, the less time the application has to compensate for irregularities in their arrival. Under typical operating conditions, this is not a serious problem for traditional non-real-time applications such as webpage downloads or e-mail exchanges. If a webpage takes a second or two to load, consumers will hardly notice. But network latency, jitter, and packet loss do matter a great deal for many of the new real-time Internet applications such as VoIP, on-line gaming, and video. Consumers will cancel their subscriptions to multiplayer gaming services if inconsistent network handling prevents their on-line characters from reacting quickly enough to surprise attacks. They will be similarly upset if, because of poor network performance, the real-time video stream for a football game freezes during a third-and-long pass into the endzone. And an interruption in a real-time telemedicine session could imperil a patient s life. Thus, while convergence is an achievement of historic proportions, it poses a fundamental engineering challenge. How can engineers structure a unified IP platform to maintain the cost-reducing efficiency of packet-switched IP networks and also assure the quality of service consumers demand for real-time services, such as voice and video, now that the signals for those services no longer travel on service-specific transmission networks? The answer cannot be that IP networks must blindly treat all packets alike by subjecting them equally to the best-efforts delivery principles used today for downloading ordinary webpages or delivering e-mails. Again, that approach would produce unacceptably poor quality for real-time applications like voice and video and would thwart the promise of convergence.
14 The answer likewise cannot be that network providers, on top of their already enormous capital investments, must so radically enlarge the capacity of their IP networks as to guarantee all packets including those associated with non-real time applications (those reasonably tolerant of latency and jitter) the same nearly instantaneous delivery needed for high-quality video services. That approach would rob IP networks of the efficiency characteristics that make Internet usage affordable. Indeed, economic studies have shown that, as IP video services escalate in popularity, this dumb network approach would raise the network costs of broadband access somewhere between $100 and $400 per subscriber. 42 Of course, one can debate the precise magnitude of these extra costs, but there can be no dispute that a dumb network approach to the exaflood would needlessly and enormously magnify those costs and that consumers would end up footing the bill. In short, the answer to the exaflood lies not only in more networks and fatter pipes, but in greater network intelligence as well, including an ability to distinguish between packets that do and those that do not need special handling to make their associated applications valuable to end users. Such network intelligence is indispensable, moreover, not just to consumer broadband networks, but also to the next generation of IP services such as videoconferencing, rural health care, telemedicine, distance learning, and highquality telecommuting capabilities. To ensure quality of service, these bandwidthintensive applications are mainly offered today over specialized virtual private networks (VPNs) operated by broadband providers on a single managed IP network. Most net neutrality advocates do not explicitly seek regulation of managed IP networks, but the breadth of their regulatory agenda raises serious questions about whether net neutrality rules would spill over to these network-based VPN applications, given that the physical infrastructure used for these managed services overlaps extensively with the physical infrastructure used for network-to-network traffic over the publicly accessible Internet. 42 See, e.g., George Ford, Thomas Koutsky & Lawrence Spiwak, The Efficiency Risk of Network Neutrality Rules, Phoenix Center Policy Bulletin No. 16 (2006) (http://papers.ssrn.com/ sol3/papers.cfm?abstract_id=925347); Richard N. Clarke, Costs of Neutral/Unmanaged IP Networks 21 (2006) (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=903433); see also Steven Pociask, Net Neutrality and the Effects on Consumers, American Consumer Institute 14(2007) (http://www.theamericanconsumer.org/ ACI%20NN%20Final.pdf).
15 The medical and disabilities communities in particular, which increasingly rely on innovative applications now available only on managed networks, have thus reacted with concern to net neutrality proposals. Dr. David Charles, Chairman of the National Alliance of Medical Researchers and Teaching Physicians, explains: [T]he medical community is taking a wary look at efforts in Washington to pass Net Neutrality legislation that would discourage further private investment in advanced broadband networks.... [T]he broadband connections required for telemedicine are not the same as the high-speed connection that brings me my e-mail. Telemedicine applications require extra broadband capacity to deliver broadcast quality video so a psychiatrist can observe the facial expressions of a patient. They need to move data fast enough to capture high-resolution images so MRI results can be evaluated online. These are premium connections that must be not only super fast but extra secure. They need priority handling by network operators and that is just the kind of handling that Net Neutrality would forbid. 43 Hofstra University professor Frank Bowe likewise warns that Americans with disabilities need, and would benefit greatly from, intelligence in the network and not just at its edges. An open and intelligent network may meet the needs of this population much better, at much lower end-user costs, than the [dumb] network sought by some net neutrality promoters. 44 As examples of bandwidth-intensive applications that require network intelligence, he cites telepresence (i.e., the remote monitoring that allows many people with disabilities to live in their own homes, away from health care facilities) and video signing. 45 Finally, to protect their users and the public interest more generally, broadband providers must differentiate among packets by filtering out those associated with viruses, worms, spyware, denial-of-service attacks, and other threats to network security. Indeed, if broadband providers were barred from performing these tasks, the Internet would likely come to a grinding halt. AT&T estimates that more than 80 percent of the e-mail bound for its network is spam and that approximately one million home computers today are 43 David Charles, M.D., The Broadband Path to Health Care Access (July 12, 2006) (http://www.medicalresearchers.org/node/49). 44 Frank G. Bowe, Net Neutrality and People with Disabilities 23 (May 2006) (http://people.hofstra.edu/faculty/frank_g_bowe/netneutralitywhitepaper.doc). 45 Id.
16 infected with bots that reach out to other computers to propagate malicious code. 46 As network threats grow more sophisticated, network owners must adopt sophisticated tools of their own to counteract them. Regulators can limit such network security measures only at the public s peril. Ultimately, there can be no serious debate about whether IP networks should remain free to differentiate among packets, because banning that practice would have catastrophic consequences for the public welfare. Even Columbia Law School professor Timothy Wu, a key exponent of net neutrality regulation, acknowledges that no one really believes in systems that ban discrimination completely, including on the Internet. 47 The only question is whether, without the slightest hint of market failure, the government should put itself in the business of prejudging the marketplace and distinguishing preemptively between good and bad differentiation and then banning the so-called bad differentiation before it happens. 48 As discussed below, the answer to that question is no. 2. How providers manage their IP networks. Having addressed the why of packet differentiation, we now discuss the how. Two overarching themes are worth noting up front. First, many of the most successful techniques for optimizing the performance of given Internet applications today occur in edge and backbone networks, not in access networks, and there is no principled reason to disfavor some, but not all, of these techniques, given their similar competitive effects. Second, to meet the complex and changing needs of their customers, providers use many network-management techniques, often in combination, and there is no single right way to manage a network or service. Applications providers and network engineers alike need all the tools in their toolbox to satisfy those preferences as quickly and efficiently as possible. At the outset, it is necessary to debunk one particularly irresponsible myth circulated by net neutrality advocates the myth that the Internet is, and has always been, a collection of dumb pipes. The SavetheInternet.com Coalition, for example, claims 46 Sarah D. Scalet, Introducing AT&T, Your Internet Security Company, CIO, May 17, 2007 (http://www.cio.com/article/110250/introducing_at_t_your_internet_security_company). 47 Keeping the Internet Neutral?: Tim Wu and Christopher Yoo Debate, 59 Fed. Commun. L.J. 575, 577 (2007). 48 Id.