`v.
`Software Rights Archive, LLC
`CASE IPR2013-00480
`
`
`
`-.......... ._.. ......
`
`.... c;-. ~a...M..._..._.._ .... ,.._ .. ~ .. c.J..A.• .......
`e..,.tc:"..._,, • .,.._.._._ ... -. T-a....c.........,mt.6o
`....._. . ....,_ ...... ns..~.c-..o.u..a. ..... ....,..M,.._
`-"' ,_ ...... c..a~aJ. ,_,_.,.. w ·~ "-'- wca Ill.
`~W · ........ ~IALnt~ ...... ~V-JlU
`Ao.mi. t ,__ ..t ,__,.._..c.._. .. lAA • ........... r.. "" ..... II
`~ <:...... r-w.cll'wll...,. Ddhi- 110 ""· ....... ,.._c-. ND.-
`~.M-...k ...... AJ.enr.~ IJIO.Nr."-W ,._,,.._p.;.,.
`l ...... I !(I • ......... ISMit. Ab-U triJ) t.i, J~ ,..,._ ""-•· ......... ,.._.
`..... Jit6.-.i1Jin
`,..._ .... Lal~ CIAo;..
`110 s-.1. a-.. 9C2I. aR.L ........
`rs.~··u...rs-J~-......
`............. ~(\IWS......,
`
`N I
`
`I r • J 4 J
`
`;
`
`tMf ........ , • ~ IWi CltlloLOUDntiiMUCDft:l un'Di .. ~
`.......... tt',..
`........... a-c .. ~--rrna. ......... .~ ......... ......_. _
`
`,... ................
`...... -.._.,.,.. -----........ ~~ ... .,.,..,..
`-0)>&-(cid:173)
`__ ,,... ....
`, -.............
`~-.. , ......... ,__,..._..__,... .... ,,.......,.._,.
`
`c - n. ..... ----""' ............................... _.,_
`........... ~ ~ .. ~war-..__ .......... -c;...p
`-. ... ~ ... ·s r....,...-.c-;.
`HIHI tf1M41M:k,J
`ISIN I " MoH 1-4 (IM.l
`c..tr""-' 1 ...... 7
`. . . . . . . . . . ~-~ . . . . . . .
`• ~ ,_...__.... .. w-... ..... ..U......S... l T*'c-.;.
`......... _
`..... ., ................. _ ...... a.r.
`,_,.. """"'-
`........... ~ -..t ~ u._. • ..,....., • .,......_ .... .._.-.
`........................... ~ .......................... .......
`~t~t~~~.,.....,_ .. ...,,.,.., ....... _ .................. ~-.4
`,.,... __ ................... ............
`n. ............................. ., ................... ., .... _
`~·..---~ ........ ......,.., .. Pk..,..O.._,_
`....... ~ . ._ . . . . . . . ,...,.. ................. fJftq J.,.
`
`PL....,
`
`,.,.... ... ~S...J ~
`S.. IIIIAMIG.-.t ... W.. .... • ~., O....t..p
`
`..,. ..,, . . . . .....,_.,_ tNftl. ~..,.., -· .................. ...,._ .......
`
`
`
`When. and How (Much)
`
`Judge of a m3n by his questions, rather rhon by hitan>wers.
`-Voltoire
`
`efurc we 11ke a long journey atOUnd the cor.roun anJ :mpli·
`or ioru of oearcb, it mak<s serue to ~ our bonr.gs. Back
`v.hcn I,..., a cub repomr. I was 12ught a> ll\JWe! 6\c c;ues(cid:173)
`about any to>piC before wrinng >bout it: who, ,..lut, where,
`anJ when. If you cr~ allSWfiS to •II those Cj'JCStions mto
`bd pangraph. then y<>u'd ess<nti211y done your job.
`Sut a> lhaoe 6ve questions I quiddy leoma! to odd • sutth- -
`howl· and • corolLa')"' who'• m:tking lhe money, •nd how much?
`Wo'll get to the money qucsrioo last, but first, I«'• "'idrw the how.
`
`So how does a -.arch eng:ne worltl 'Then:'s a ''CrY· "Crf long an~owcr to
`tlus queHion, but I'll stick to a shoncr one. In cstc~. a surch engiiM:
`connect< ,...,rds you ~rocr (queries) co a da<alnse illw aeated of Web
`pages (an indc.l. It then produas altst of URU (and surr.nwics of
`conttnt) it believes Are moa rdaanc for your query. While mere lUI!
`
`
`
`cxperi mmul approacbn m se2tc:h tluu arc ncx driven by Ibis pu.t·
`dsgm. lOr the most pan cvrry major sorch enpnc u clmca by !his
`teJC·b.ued •pprooch.
`A sorch cnpnr ooculsts of tbta: nujor picco-tlx a-M, d:c In(cid:173)
`dex. 1nd the runame l)"tml or query processor. which is me ln:u(cid:173)
`r- and rd.rr.J softwln' that coonccu • wcr'a querict tO the iC>da.
`The runtime aysrcm also ~ the call..mpcram quauoru of rd(cid:173)
`~ md rankin&- All lhrc< poec.cs ate inrq;r;U to the quahry and
`sp=l of tlx mgiM, and there arc litcnllf hundrcda of f=ora in
`t>c:h that a!Tcct the o~U ..,.,J, apenmt:E ddovcaf. llut dv: 1mb
`ue prcay much lhc umr fot aU the cnpna. A> 1un Utav. a search p(cid:173)
`oo= now at Sun Moaosymms, puu 1 on hiS cnxllc:u luics "On
`Scarth." "The r .... of the matltr IJ that rlu.re reaDy ham't been mu.:h
`progress m rhc baslC ~of how 10 aarcb ainox tlx ~C>ello. •
`The search all oraru with you: y<>or query. your uotcm-cl.e de·
`un: 10 !" an aruwu, fuxl a •••e. or kam aomcthiaa new. lruCtU
`droves ae&Kb-• mutm rube rcpeatlng time and agam throu£houc
`this book. W~ II tctlmo the query a btt mate in the: "What" a«tl011
`below, but on ·~~ ~ <IIIU one or two shon -..ords 11110 a quar
`bo1 cadi time we ldrth, and we did on •n avc:n&c of twv or ao re(cid:173)
`sults amonc the miUtoru an cnpM often lisu. In :addition, the nu·
`age Web scan: her oonducu ai>Oilt one scan:h a d.y. Of tounc, that s
`an ·~A amall pcrtt=gr of hopdosly conna:ta.l awfcn COD•
`dua hw..Utds of acarchct a d•y, and many marc do no marc than
`one or two a month. (All tl~ ligures, u one mtglu etpot."l. arc
`powins .,.... tunr.)
`The pro.;ns of hnw .... 'C s•r our mula lttltJ ·""' •he cnwkr
`The cnwlct u a ~ullicd ool' cwarc program thac hopa fr, m link co
`lmk on the World Wide \\'rb, scrrfing up the paces n finds and
`Knding them l»d 10 be mdcxcd. It'• SC"ductlvc ro rhlnk .,f uawkn
`as tlnyliclc toboa "'2ndering the wut halls of c:ybmpaa:, but tile
`truth is .. bit 111<1Ct mund.u>O. Crawlers are m r.cc horndx>dta, lit·
`tint; on their own ~rs and ~tndiag our v:st numbat of mpusu
`ro pages on rhc lnccrnct, much u your bruwsc:r does.
`
`
`
`Who. Wh41. Whar Wby \\lm ADd Hem J"'!l'bl 2J
`n,-~ brinJ back Web pqo. wluch dw ano-ia thm
`oii to the inckur h abo nW note of any linU 11 hat
`on tbc P'lt• and qucua thoae hnL In lu rcqucl lik(cid:173)
`out yet IDOC't rrquau to the IIC'Wiy found W.la, "'-hlcb lind
`&tid 10 on. ad tnlinitum. ~ the OC!VJC:t be
`linh
`cnwtcn u com;mx, ..h.otlhq do b Ptttl"f
`lc tbcJ 10 otT
`a cndleu biJI&t of d~ fot UIU.a and chq rcpon back "'hot
`found Cnwlco ruw long betn the len ""ble of the
`cnpnc'a wm~nu, but they an:~ the cnotltmpot
`The IIIOfC. to thty Cf"l10i, and the lllOR Cnquuttlylhey ~1
`the mote c:cmpku the lnda lt. Wftm the indo Is mote
`10mplcl:t, tbc aarcb rollin Fit" (SERPa that arc rn:orned (Of 1
`it..rtiicul,ar quur han • pater chance of bcl:!a reb-am
`Wv vmtom d cr.wlcn d~red and tndaat only the urlu
`Web paqo. but tudor'• mon advanad YUt~orn lndn the con
`of the C!Ulre 'Wtb ~"£Co IU wdJ II manY d:Ifmn& f.lc l,pa NCfl
`Mob< Aaobor (PDF). Mta010f1 ori>CC docummo. aw~.o and
`and tYCn 11w tptttl1c me:ad ua-wuaun:d ift!omutJOn proo
`br • •• owncn about the ~ or lnfarmatton bc~t~~: cnwkd
`Th aawlu ocncb Ia duo bad ro a IIUJ;lltt c!a hue a!Jcd the
`11K indo braL lmo IIC"Ctal p!«O. ckpmdins on ...tlnhc-t
`dau lw bern pro cued and nude rudy ror axuumpooa by
`..-•ncactalike J<lU &tid mr. R.aw lndna ~ nthct ke luu orp
`by donuln fot any pnn a rc, the lnda .. uJ
`t aU the P"£0
`rlutlltt, •• -u as all pcrttnrr.t &n!otmuioo abot."t tb sc ~
`the words on the pqe. the h~>u, the ard:cf tt.Jt tell •round and
`•b n a Url.), and 10 on. 1llc: onfornuuon tJ 01p
`n auch a
`rlut tf you • - the URJ. you an lind the ..-onb dut ..... rr
`to rlut URI..
`\l:ny u thulmratwul Beau~ me nm 11cp In antJt1i a unan
`u to lnvm the daw.--m """"'t, ro auh •
`o( watdt
`lli'C thm "'OOa:cd ..ub l1RLo.. So wbm JOU I7JlC •ou:a Moa
`IJI:O &1Grd. bos, thc cnpM ~ o(
`the URI• th.u lnducle thole 1"0fdt.
`
`
`
`The first engines on t~ \t'eb CS$Mtially =cd to tiW point,
`and not much funhCJ', But sinct the late l990s, the tndex ha.s be·
`come a <ignilicant ama of in!IOV2.tion for aU Je2lCb compwi.(cid:173)
`wbc« much of a <carch engine's $CCtct ..,uce is applied.
`Tlunk of the index as a bu~e databa..c: of imponanr informa·
`tion about Web lites. Innovative compa01es like Googl< hue m:ade
`their reputation by srudymg that database-noting •tatistieal p•t·
`terns and algorithmic potenttals, divimng new ways to kvmtgc it
`toward the ultimate goal of providing you with more relevant re·
`aulu for your queries.
`The proe<:s> of grokklng t~ index ls n:ferred to as anal>""·
`('.ooglis PogeRank algorithm is an example of analys"' it loola ar
`the lin Ia on a poge, the anchor text around those links, and the pop(cid:173)
`ularit;• of the !'"I>CS that link to anO<bcr page and facrort them to·
`gcthu to d<ruminc the ultimate rdcvancc of a parncul"' p:og< to
`your quezy. (While Page!Unk is often understood to be on "aJ(.
`knowio&" .Jgonthm, Coogle, in f.act, lookt at more than one hun·
`dted f2ctors to determine asice's relevance to your keywords.)
`Through the process of analy>u. indexes an: populated wUh tags.
`anOther kind of met:adam-data about dato. Page• mtght be taggcJ,...
`wriiiCIItn • cenatn bngu.·,:, for cnmpk, or as belonging to a cen:ain
`group such as porn,spom, or 1\\tely updated. Thi• muadata is cririeol
`to an engine's ability to offer you the most rdevant resulu.
`Once the crawl dara is an•ly.e<!, index..!, and ~. tt's
`dump<d into what's c:alkd • runtime index~ dambasc rcody to
`,.., .• results to Ulen. The runtime index fornu !IOinCthing of a
`bridg< berwecn the bock end of an engine (ics crawl and index) ~nd
`•~ front tnd ("m qucty server and uso:r interf2ce)
`The query server is sofcw.rc th•••nnspom a user's <earth query
`from the user LMerfxc-<], bome pogc of searclt.yahoo.com. for
`10 the runtime index, then .ohutdcs SERPs bock 10 the in·
`ownple-
`t<rfacr \t'hile much of w engmc'• mrdlir;encc iJ b.i!t inro mal)'!is,
`the query server CUI hold quite a bit as wdl. If )"KK'n: 'PC"' any time
`banging on dif!i:rcnt typeS of search engines, you can see 90tne of thor
`
`
`
`Zl
`Ibc Sctnia
`to C"'alc rhc \l:orld Wide Web.' And "wu Utry J'a&c and 5c:gcy
`Bnn'o ar:mnpa co unprow: Bc~L«'o World Wade \l:'tb chat lcd
`to Goosk. 1bt neodlr that thJOO!. thac clrorto toged>tt u o ra·
`r10n-tho p...rocc of po1nuns co ocher pcoplt'• work Ul or<kt 10
`butkf up )'WI own
`Whir:h brinp w b.ck to tht oripnal rar.arch l'ogt did on bad·
`linh. Hr rosoncd tlur the enure Web wao Joo.cly booed on tht
`pmnuc of anuon .111d wnooation-Ftn all, wbat-.ns a link bur I
`CICIDOn, and what wu the co• dnaibJnc thar!Wt but annoatlonl
`If he rouLl divine a mcthoclto ClOUtll and q<Wify each bockhnk on
`the Web, u Pace puuot, "the Web ....,..ld btoomc a more nluable
`place."
`"In aiiCDIC, • POF cantioun "tho Web u this: anyone a11 atiJlOo
`we anytlu"' ~I)' raDiy Juit by lonk.ing to u. But tl>t •my vnuocu
`of hYJ>Uta<t had a '~'>~"" IU_,ou couldn 1 follow linb lA tht
`ocher duca10n &.c!UWb wu obout ~ that. It oecmal kind
`o( cool m pthcr all !hclinko on the Web and ~nc them •
`
`r-. bypothaiud 8adtRub. u he c~llcd hu proJ«t, u a 1)'1tcm
`
`that would ditcoftr linlu on tht Wtb, 110"' them for arulyW. thtn
`tqoubUah them on • way char made U pnalblc lOt anyone 10 ~ee wbo
`wu linkins to :any coom page on the Wtb. An ltllbttJouooda on any
`oalc. buc J>asc dodn 't JCI out to nuke Bad!Wb wotk on a amall~ct
`ol rae JlGIH. lrutai. be tbouglu b11= wiry noc 10M the problem all
`ar once, for llw ,., r..u WiM WHo!
`To Wldm.W: .u<h a wk mp11a a rarha oudadow bene
`While l'lse WU 11011118 )WI !he hnl:o not !he conccnts o( tho cmm
`Web-he ocill had to enol the crultc 'IICb to 6nd !hoot linh. In
`199~. ouch a IUt wu quire ~ ,
`At the urn. Pose con<rtvcd ol BedtRub. tho Web comptiJcd an
`a!Jmaccd I 0 mdllon docwncnu.. with an untold number of liAb
`btrwCCD than J>asc li~ that ic wu aon.....btrcln tho 11111" o(
`I 00 million. The number rumed our to be much larscr. And
`lonsu l'lse w:aued ro F naned, the bigu the Web became. In
`cat!r .Uyo, tho Web wu pow>ng :u a nnr of more chan 2,000
`
`
`
`Crn;'c b Bqm Zl
`
`ccm a JUI· The cornputmC mQUittl mjUltcd ro ~ aadl a buJt
`~ ~ t..,.-1 me lmUl bg.,ncl, o{ a mwlrm projca. Sowewlu!
`wuwm o( wht: be ,.... s=n1 uwo, ~ bqa%1 ~ our hla
`cnwoicr.
`The ida's ~ a:>d Kale lured Sctu7 8rin. llrin. 1 p:.ly(cid:173)
`awh who bl )" ;w.i from proj«t 10 ptoJ«t without lftt!lnJ an a
`mcm ropo:.' fouDd th. prmmr bd1md ~ E.ucW
`·a
`talhd co loa o{ march poupo. II'OUftd thr ochool. 11M I'CQJ ...
`"and mb ..... me -
`aaun p:oj((t. boch beaux k nd-!M me
`\reb, which rc;msc=s hum.n ~ and bca= I Ucd &.u.,.
`and me CJ<ha rwo people who~~ wU 1P •
`The rwo cxb<n ~ db ~ a:>d 8riA wm: Sox! Hamn
`and Al.tn Scmmbrta. pdum •llhana •ho lud ban uo!ptd CO
`d>r projca. (Each PhD and<ch~ ..-.u ~~ an•IO!•uul a~
`umolly • cw='o owdc:a lor! "' 10 nW.t a l.!nlt CUD 111M17) tw
`and ScooubtlJ ended up ~ r- chc ~ bdOR
`-
`Coop, rully COol. olf 0.. com IIPt ~ llecla oumd ouettt~
`fut lnttma mnpr.Jn.. I W.O.. wmr an 10 l'nwld ~
`Lani• brocbtt, Cut ~and tr.a dd k 10'
`(a, moft chan
`SSOO milbon ~had almdy blinchrd 11w \la:ha Undn
`pound. 1 popubr ~ .m, ..tulc an~ ac M~~
`and mil rum amdq
`
`lbc AuJ.uio· of R.ank
`~ lllld me chat 11 hod NWt be= his ll"""'don 10 autt a oniCh
`mpnr-tndmi, be and Brin lud 110 Idea what uacful ch!np lhc
`pro~:c1 m!p rum up Bill In ankr 10 Cln!r 1.\ockRub.lhq hood to
`the .b In Mudllm, PoJr powud a••ld at
`-
`hit '*11 ' - f'l£t 11 Su:Uor:! (moa CS pad otiadtnn bad
`ac>--.:~ Ia • lOOK. The UPiet ...n:d ourwud Cram Ibm
`
`lbe loamy o{ me \t"d>-oo -= •bm 7W owt. comw
`
`!'OU'D SCI )Uar abour Ud1"bot ~ mm u ID p
`,Q,IWiJI~ dJC CIXJtC 'l'l"cb co ~ dar IUitl o{ ill L::L ...
`
`
`
`dcnUt P'~ishj ... d he*
`
`I
`
`W1dcnal:ing. bur .... awlioc _ _ .._ 8eclr!l,ub'1 aue uu.
`
`.. lion lay.PJF-....,.-,. .... CIIIICifl olllllllda& ia(cid:173)
`groph ...,..Jd ~ ... juiC •-laldaa 10 .me.. .. -
`I . . abe- rl die .....
`ald(cid:173)
`cdy. the itryo -of wllo JialleciiDwllom. '-d 011 -. . . .
`,_ r:I the lilc dw- .............. 7 ~ .-1 allilr, ._
`.__. the .... lix
`... .. . ,
`aara..--dle aadlar -
`aitical ift de~a~~W~ia&nnklllllu:lau c
`If BackRub Jute.... mew...
`ol a . . ltcoald p elm sb
`a ftllatiYe ranJUns. For .ny p.-. -coa~c~- aacaal)' wbo(cid:173)
`linlcins to r1w •·· bur the_..,,.._ liDb u .a. Caainly
`::Jw
`-would bo usoful, .... ••
`dftdcpmtlll ~-They &.ID't,. dedcW .. lhae-.(cid:173)
`8cins uodiJI - Ul
`' % ...,._ upc ,. , . ...
`Brio's raoardl (and hu &ea..a-IDr .UofGoap's pra t, ..
`pany iD llod<Rub. bur me......, «T•--&r &-, . ••
`cally real and oruaaliato" ......... "JJur dlae io DO -If,.
`... doins oadaalc -m ID,... • dllap m. a
`
`mind. "Mypla-n>wmkllll W tdlor-.ldbandeml.
`
`imprri sl
`Lc ,. ' '
`There on: p1<ary o1 m
`thot .,. silO p!'ll:llcaL 1
`Milled bath. and I dida't lhiak thea -IIIIlCh of a tndHff' 10 be
`modo. I fituml if I cadeol .. !w'd'• -mine chac-.... 10
`P'*•iolly btDefil a Joe ol people • , , dial I would ba opeiiiD -
`mcrdaliziag ic-t01hat 1-Jdn't bo IIJae Tala."
`Once Paac and Brill W awW die Web and IIIINd a paph ol
`iulbW,Ihq~tof Dllwa=-kmsmedtod lw lupnd
`fli Ulllll 10 I
`' thK I ._ -
`f>18C would boa Ulef'ul l'*lraa ._ pap'• rmk. He silO tbeoriad
`mar-'> link ncecled .. - ~ bued on !he l.iak- of ill
`fi
`originarins PilL But JUda •
`h aeua 1 clillicWr and
`)'Oil no1 oaJy baoero CIOWif a pa1rkuo
`1M mtrhanorial d.•P 1
`1u ~""&'~:'• IUW. ,_ allo bawl 10 aJUat !he links anricd 10
`li.W. Very qulddy, !he mllh pa nther camplicalcd.
`Fonurwely,llrirl'1 pra ';II us pr..ia l1lllhautica cauld
`
`by citalillll anal,. Pip 6
`
`
`
`<i<xll!k ($ !IQm 11.
`plic.l to the: problem. Brin, the Ruui:an-bom son of a NASA ocien·
`mt (hi.s nvxh<r) and :. wtivcrsity math prof=or (hi.t fathu), emi(cid:173)
`grotcd to the United States with hiJ fam1ly at the •gc of si.L By the
`ume he was a midd(e.,chooJ.,r in iuburb<Ul ~:.ryland, Brin wu a
`rccosn=t matb prodiAY. He dro~ ou: of lUsh achool a yeu
`cufy tO enroll at the Univastty of Maryland. wbc:c his father
`t•ughc. Or>ec he sraduared be immtdiucly cnrdled at Sunford.
`where lW mlenu allowed him to goof off. The W1:athcr w.u .t<> good.
`he told me, that he took monJy nonacademiC dasset-a.tling.owim·
`ming. diving. He foeu>cd his inldleaU<Ol energies on incen:sting
`projecu ruher th.111aauil ~
`T ~r. Page and Brin aeared a r.ankias ')'Item rcwuding
`link. that CliTlC from 10utta tlut wuc imporu.nr, and penalizing
`those that did nor. For example, many sites link ro ibm cam. Those
`lieu mighr range (rom a bwiness putnct in the <cchooloAY indus·
`uy-lrud. p<rhar-to a tcmage progrunma in suburban !Uinoi.t
`who ~nked tO IBM bccawc: be Just p a n...., compuur f01 Christ·
`mas. How might .Ill algorithm clcttmune rank bm.un these two d(cid:173)
`tlltions? For a human obsci"'Cr, the bwiness partn"r is a more
`important link, in 1er111.1 of undem&~~ding IBM's plae< 1n the world.
`But bow might an alpitbm undmtand thu f•ct!
`Pa&c a.'ld Bnn '• brc:olahrough w.u to <=<c an aJcorithm(cid:173)
`dubbcd l'a&c!Wlk am Page-chat maruges ro take inro =owu
`both the numb<r of linlu into a particuLar sire, and the numb<r of
`link$ into each of the linking sires. This mirrored rhe rough ap·
`pro..b of owlernic citation coundng. and as it turn<d OUt, it
`,...,rked.ln the IBM rumple abo~ lei's WWIII: thu only a few sita
`link.td to the <cenager's ate. Lft'• furthct :usumc the """ that link
`to t!-w: =•g..r's arc 11miluly b<mt of lmks. In CQ!Itnut, th<lWillds
`link to lntd, and those aiteJ, on avtroge. also havt th<lWillds
`lites linking to them. Under 1'18<R.anlt, the teenager's lite would
`u lc:u irnporulll t!un • slrr ~ ltlld. In Ibis cumpk, then,
`Btin's nnk.iasmetbodclogywould judge lntd as mon:im(cid:173)
`than • auburban teenager-.. least in relation to IBM.
`
`
`
`B De '
`t
`1\lolu, sf' • ._.,.__ . . ~..tllriaWIO
`....... = ...... ...
`__ _,......,.vi •I
`thoi· _ _...lior. IIIII loa popular ..... feU .-ud me.....,__
`papoalot sites ..... 10 -
`and the """' vi it ... this: -
`.... vi
`"' mer &ddlod ...... • nsu~a ICtiii'DCd by .-........, Brin ....
`..... rallud mer .... ,. .. onto -m•ns mat might .,_ implica:
`lions 1\w lntcmct acarch. In faa, me ida of applying &d<Rub's
`rolred .,.., raulu 10 sarch wu 10 natwal. Page recalb. that it
`dlch't ....... occur 10 tbrm that mer hlciiiUilde m. leap."' it .....
`Bad<Rub alrady worlc<d like a acarch <t~sin<-)'OU s- it 1 URL
`and 11 5""' )'OU a list oi baddinks ronkcd by i~- ~ real·
`iud that ...., had a queryin« rani. • .,.., Dnkinc !hat wv wcf..l for
`a let of thiap." Page recall<d. "It 5""' you 1 goOd .,....a ruking oi
`pops and ord<riD& of IOIIow-up .,....·
`..... and Brin quiddy nociad that Bad<Rub's raulu _,superior
`to :hole of uaditional a<at<h cngin<t lilcc AltaV'IIU and ~. which
`of1rn rcrumod im:bont resulu. ~ thoup;hr. Why •rr tht] rrnmriltf
`tMt rmJis tNt - D/Mo...lf n« im,.-t>" Page roaJh. '"Th:y ......,
`only looking at tat and no< cocuicltring tlu• Other 5ignal Once you
`ha .. it. it's P"'"Y obvious mat this . . . is useful in ...,m.·
`'The tip>al-- bctlcr .._, u Pagdtaalt-beamr me
`wbnha
`foondation of Coogle's numod ...:m uucc. To -
`... Rank worlc<d wdl in • sarch application. llrin and .,. hacbd
`tOF'hcr a Bad<Rub acarch rool. It san:hod only the words in URL
`rllla and oppliod PageRank to rank tl~ rnulu for ~knnc<. biiC in
`r<RJIIS were so f.ar superior to tradllional acarch cngin..-whidl
`rank<d m<Ktly on keywords only-that Page and Brin knew chcy
`
`w:rr -o something big. 1
`
`And nor only wu r~ cnr;in< s-J• Page and Brin roliud ct
`...,..Jd oa1c .. me Web oa~c.t-PageRank worhd by anal!rtilll'
`li:llu. 10 the bigu me Web r;oo.me bcmr me cnpnc ....,..ld be..
`fw lrufirod rbe fiJOIDlkrs to IWIIt their new cnr;ioc Goor;l<.
`~.me term l'o<rbc number I f.,u.,...d by 100 umes. TL-""
`
`
`
`C~lcbRnm 11
`k:ucd the lim Ymion of GOOj!k on ohe Sunford Web s.re on Augun
`1996.
`Among a snull " ' of Sunford tnliders, Coogle was • hio. l'ner(cid:173)
`giud, Brin lUlU P•s• bcg:on improvinG ohc ,.rvi«, adding full-t<Xr
`~~:arch and more and more page> to the index. But search engines re(cid:173)
`quire an crmnrdinary amount of computing resoun:cs. Graduate
`otudenu usu•llr t.ck the rnoocy to buy new compurer<. Page and
`1\rin v.ere no cxcrptu•h. lnsrt>d they heg&cd and borrov.cd Coogle
`into W~- a hud drive from tht network l~b, an idle CPU
`from <he CS lo.ding docks. U<ing P•~'< dorm room llS a machine
`Lob. •hey f.nhoonrd. cornpuwiolul rr.nkenstein from •pare pans.
`•hrn jxkrd 1he whole thins imo SI.Uifonl"s brOildband campus ntt(cid:173)
`...,.t._ Afu:r tilling Pas•'> room wi1h <quopmenl, the youns >tudems
`con\·cned Br1n \ R'l()m 1n1o an offict' anc..l programming ctmtr.
`Hector Garu•·Mllhna, anothrr f.>ruhy adviser on th< projeCt,
`lent the nudcnu a ~un Ultra, a powcrfulcomputtt <h.t 1'4e r=lls
`hod cen time< rhe memnry of a typoaii'C. Bux IM"n th>t w:u not
`nearly <JW>ugh .-\c'JUmng ,._,,= r ... their work beamc n<arl)'.
`full-lime job. "We l~ed hardware and connectivit)' fmm a<h<r
`groups-we hod 10 much d;ua co move .round," Ry> Page. ·we had
`10 "" chem tfl open up <heir wiring Jooets, lend us chdr l'chernet
`atds."
`An early .-rna from Lury I'~'" rerry Winograd diU>trates the
`lund of problaru •he found<n encuunt<red. Dated July I~. 19%, 11
`rt.da on p•m
`
`I am almotlt out ol cNk space I 1-wrJe downloaded
`lboul •• 24 ndon tnq1>e URls.lnd ~ \00 mdlion
`• I umlt I wll ~ 111>0u18 gogs mora 1o 11ore
`In'<&.
`~ Cutrenl retat pncee are about SIOOCI'4
`gigs. • •• I ha.e 0<1ly about 15% ot !he pages built seoma
`ve<y P!OfTiiSil'lO
`
`
`
`Ill! Ih~ s,.,m
`
`blind judgment r<g:~rding their site. After all, thll was the lint time
`anyone had claimed to rank the inherent value of a Web sire. As it
`still doc1 to this day, such jud~nt CToked a powerful rt$ponse.
`"I am in a sate of stupdiearion rcprding )OUr ranking pl'OCDS, •
`read one typical compbim. "Since you ::reated chis search engine, you
`need to correct a hidoous error which i; both laughable yer hurtful to
`a w..bma<ter. Plea.., type the word< 'lJI~-nes S Grant' intO your sea.rcb
`engine and look ar rhe rerula. My -..cbsite, 'The Ulysses S. Grwt
`Home Page' was vorcd '11le Besr Civil War Webme' in rhe february
`1998 imoc of Civillf11r Timt 11/rutmttd. ... This website is graded as
`the prcmocr \\cbsitc on any Ci-111 War ~rsonality or banle. Bill GateS
`C1-"CJ1 wnx~ mea pason.;~l ~mail pr~i''"' it.. ... You ... rank other. in(cid:173)
`ferior (nnd •orne pathetic) sites higher .... This is an injustice of such
`magnio u.le thar ir beg. explanation. l fed confident tltar if you take 5
`minure> to look at my website you will rank it higher."
`Pagt' and Brin had clearly hir a nerve, not ju>t with Civil War afi(cid:173)
`cionados, bur with every person who labored over a Web Site. To
`many, unleashing a ranking system based on a bloodless algorithm
`fdr like a suprtme act of arrugan..e-who were these kid. from
`Sranford, telling the world how we ranked? \\'hat did they know
`abou[ tht' work and pJ\Sion that wcm inco o11r sites?
`Well, in crurh, Pase and Brin rrade no claim ro such knowl(cid:173)
`ed~. As the><: early complaint> illu<trate, the Google S<Orvice made
`no pmcnsions of actually reading a panicular Site, or of under(cid:173)
`standing its content. It simply bid bare the often ugly 1tt11h of
`how well connected a <ite happened ro be. No marrer how gr.,at a
`site might look, or how many awards it might r<ceive, if it was nat
`linked ro by other ~it<<-ideally, sires that w<rc themsdve' well
`linked-then, in Goosle's estimati<•n, it didn't really exist. That
`cold, hard fact was hard for many tc swallow.
`A May 1998 e-mail from Winograd to Brin about the com(cid:173)
`plaints foreshadowed the power Google would soon have over nearly
`<"V<ry sit< on rhe Web:
`
`
`
`Goo&~< h Born a!
`
`Long ago, Larry came to me and was eager to do re·
`search by putting a service inlo general use on lhe web.
`I was skeptical because " opens you up 10 random has(cid:173)
`sles, with lhe number of hassleS Proport<>nallo the num·
`ber of people aMecled by your service. We ha\19 now
`crossed lhal fino, end are In lhe position where slopping
`the service w.l create a large number of complaints as
`well. But I guess thai is jusllhe cost of doing bosl'lessl
`
`While P:.ge an.! Brin didn't ~ it at the time, their carl>· rank·
`ing 5y\l<nl w:a< erching the uxes of ~n cncirdy new ecology. "" eu>l·
`ogy shaped by million~ of deci•ion• and million• of Webmmers,
`each one of them wnhongsimply to r.mk better in the Googlc inda.9
`
`A CompA~n)' Emerges
`ru Brin and Page continued expcrimcnring with search, &ckRub
`and ir< Gnoglc impl<mcontarion were g:oining bun. both on th< Stan·
`ford campus and within the cloasrered world of academic Web re·
`search. Onc puson who had heard of Page and llrin's work w.ts the
`aforementioned Jon Kleinberg. then a researcher at IBM's Almaden
`center tn San Jose, now a professor at Cornell. Kldnlxrg's hubs-and·
`autboritic. •pproach to ranking the Web is pcrh•ps the second most
`famous approach ro search after Page Rank. 10
`Back in the summer of 1997, Klcinbcrg visired P>gc ~t Sr:lnford
`to com part nores on search. Kleinberg had complered an early draft
`of his seminai"Authori<>tive Sources" paper, and P•ge showed him
`an early working version of Google running on the makc.,hifr sys·
`<em he and Brin had cobbled togcth<r. Kleinberg encouraged Page
`publish an •cadcmic paper on !'age Rank.
`Bur in the course of his convt"ation with Kleinberg, Page told
`Nemoerr. thar he w:as wary of publishing. The reason? "l ie w.tS
`
`
`
`az •tv
`couc:aawd char •• 111 •M' miPr sraJ his ideu, • JCirinberJ rold me. lr
`- PIF'• Tala c:an8icr ar work: on rbe one band Paae rapecred
`aod putidpaed ill rbe IC8dcmic rndirion of sharing resardt
`chroush published papers. bur he was also in8uenced by rbe i1IIIR
`clorcd, defensive posrure of a corporarion pRK«tins ia inrdkcrual
`properry. W'nh PagcRank. "£Patel far lih he had rbe ~ecm for(cid:173)
`mula, • Kleinberg rold noe. "lr did ~CaD a billlJ'aiiF ar rhe rime. •
`Academic &- ulrimardy won our over rhe proprierary im·
`puhe. By rhe end of thtir conversalion, the pair agreed to cite each
`orher in rheir paptrs. In early 1998, Page submin.d his 6m paptr.
`an overview of rhe l'ascRank algorithm, to rht SpccUJ lnttrm
`Group on Information Rmienl of rbe AssDCiarion f ... Compuring
`Machinery (SIGIR-ACM). Bur the paptrwas rtjecttd. Ont pttr rr(cid:173)
`viewer wroct of rhe paptr, "I found rhe overall prrsenrarion dis(cid:173)
`jointed. .. • This no:tds ro focus more on rhc IR issues and 1~ on
`web analysis. • Page ntverrheless perstvered. and the paper was ulu(cid:173)
`INtdy published in conjunction wirh a Sranford digiral libraries
`projecr.
`Shortly rhereafrtr P~ and Brin published a paper on Google
`iaclf. Thar paper. "Tht Aluromy of a urge-Scale Hypeno:tual
`Web Seatch Engine:," has become rhe moor wtddy cited search(cid:173)
`rdared publicarion in rhe world. Given the ultimate succm of
`Googlt iuclf, ir ~tems Pagt and Brin had rhar academiC cake and
`goc ro ear ir, roo.
`Back in rhe early~· Paae and Brin wtren'r surr rhey wanred
`ro go rhrough rhe rrav:ails of surring and running a company. Our(cid:173)
`tog Page's 6m year ar Stanford, his farber had dted, and friends re(cid:173)
`call tlutt Page vitwed finishing his PhD as somcthtng of a rnhua: ro
`ha father'• lift. Given hu own academte upbnngmg. Bnn, roo, was
`rducrant ro kave rhe program. Brin r:calls speaking wirh his ad(cid:173)
`viJcr, who told him, "Loolt, if this Googlc rhing pans our,
`grot. If nor, you CUI mum to graduaa: school and finUh your
`SIS." Brin chudkd,rhen added· "I said. 'Yeah, OK. why nor? I'll
`give It a rry.' •
`
`
`
`Through Stanford, Page and Brin had access ro an extraordi nary
`network of Silicon Valley bu<in<M imdligcncc, and by 1997, Page's
`brother Carl was already hard at work building eG roups. The con·
`sensus view held that wre were already • g:tSSic of •carcb-related
`businesses, all well funded and thriving. Yahoo. Excire, AII3Vista,
`I nfoscek. Wired Oigit3l's HotBot: the li>t was long and growing.
`Page and Brin rrasoned that the best course mtght be to licen~e their
`new technology to another company.
`The inventors f.1eed a classic entrepreneuri~ l dllem ma: if rhey
`started :a company, it could be cru<hed by l:arg.:r, richer compctiror..
`On the other hand, of the ~;ompany rook ofT and became best of
`breed, the upside would be huge Yahoo, Excite, and others already
`had multi-hundred-million-dollar valuations. But taking rhen1 on
`~ risky. Page and 1\rin chooe a more conscmmve course. Bc:tter to
`licen<e the technology to a major pla)•er, they reasoned, and avoid
`the ris~ of a <r.m-up.
`Tite first mempt to liccn<e Coogle's technology occurred very
`early in rhe projta's life. Vinod Khosla, the "ell-connected partner
`at the venture capital firm Kleiner Pakins CaulieJd ~ Byers, had
`lc:amed of Coogle through hi> own Stanford connections. I m·
`the
`presstd, he tried to persuade a cnmpanv he had invurcd in-
`n .... -lr publ.:: P.xdre-ro acqmre the technology and its crcarors'
`scrvi<es. This incited a tlurry of e-mail lxtwccn Khosb. Page,
`WID<l(;Dd, and Brin. Page"'' the price for Coogle at $1.6 n111lion.
`Kho•l• said he thought he could per<ua<lc Excue co ofT~r $750.000.
`Rrviewmg thor t:arly e-maol exch2nge>, u'• rcnwkablc to ~••
`Page's tnctptcnt bwinos sawy. lie knC\' chat Excite was m heated
`battle wuh the much larger Yahoo, and \JW Coogle'• technology ns a
`to Excite's gaining a compctu•vc edge. Waon'r that wonh bridg
`tht dtlfaence between hi• price and Khosla's oounrer offer?
`market leader usually is 21 lean ~ times as btg u the number
`Page wrote to Khmla. a ''cteran deal maker. "[Coogle's] sig(cid:173)
`ltiantly unprovtd acarch technology will help F xcite gain and
`~!nuin market share."
`
`
`
`Pace abo arped dw rbete would be • sipWiano: aJOt to Excite
`sbculd his rechnology end up ebewhere, bur the Esciu exccuma
`were unconvinced. Khosla visited the &cite campus 111 persuade the
`CEO, Gcup BelL to cbanae hia miDd. (Bdl. a -_.,.! publishing
`aecuriw, ...,cja....t the "edult lllpCIVilioa" brought in by Excite's
`inveaton.) "Beel ducw me out of the oftice," Khosla lllld me wilh a
`wan smile. • At last I uicd. •
`o- the-..... of the """' eigh....., months. the young inven(cid:173)
`ton gaw clemonatntions of Goop to natly every oarch company
`in the Vallq, from Yahoo to mr-eiL 'J'ho,y aloo llw cd their oecb(cid:173)
`lltlitv 1D - . J - cwpitalim h7- fouad their wbnoi(cid:173)
`Off ...... lu ada-dacpd ICUCiena on their way. "I del
`led S.W:ICiad!. fi• :lrr af dac(cid:173)
`them 10 10 Jllllllllhaad," a
`I Jerry V..SIIIIII DaM Pilo, dac , .......... af
`defima para! H
`Ylhoo. wac more~ 1M tbq, too, took a JIUI·
`becoming ponals.. Pace __. .. of the companies
`"They -
`pw ...
`be visired. """pcablbly would ..... liccnecd it j( . . -
`dac _ , •••• (lui] me,- - inaueacl in MUCh.
`"'J"bq did haft hocolcopa. thouah.. he adcla clrily.
`Suflice it to ay, __..--hlp of mind for-._
`thu- ·a-~-• . Aad UIJWIIY• ill the._ 19901 the p i (cid:173)
`
`... in dac - I~ SaJda -
`
`nor to ~end people -.y from 7'l"' portal, as RaJCh did. I t - to
`~ them there.
`Rqcctcd buc not deurml. Brio and Page- back to Stanlixd
`and bpc warldas 011 Gooa1e. which they bpt up and runnins at
`~· "'A llid ID CIIUICha, '\Ve dao't t:a~e,'"
`or_,.. it'l jail Ill!.- 1
`.,.. "''1-"•it-- M&Jbeil'll-latoaa~~...-rq
`~lt.' •
`------
`111M: by dac middle of 1m. ..... _ ........ -
`lallindell l'lp of m. lwht:/1 cGnlupl t.•- "It -
`
`I
`
`a CDiiUIIOCiity-- faauc
`
`
`
`.G~X>:~I< h !lo Dl 6.1
`
`By late 1998, Google was serving more chan ten rhous•nd
`qu.ries a d~y. ~ or was dear to P "!;< and Brin thot tht <cf'VICc
`would quickly outgrow cheir nbiliry to beg resources to support it.
`St<trting a c:ooap;~ny became the only viilile alternative. The founders
`turned ro anOther faculry adviser, David Cheriton. Cheri<on, who
`beads Stanford'• Du~ributed Sptcnu Group, was an old hand at
`company form>rion. He h•d founded Gro.nite Systems, a develo~r
`of netwOrking technology tha< was sold to Cisco Systems in 1996
`for $220 million. Cheriton • ugg<"ted that Brin and Page meet wirb
`Andy Bechro!.hdm, a founder of Sun who was active in early·"age
`invc:sonc:nrs.
`As Page recalls, Brin sent Bechtolshcim an e-matllarc one nigbt
`requt"Sting a si·-down, and Beduohhcim answered imm<dia<dy. He
`susseS<ed meecing the next morning at eight o'clock. •n hour a<
`which <he graduau: •rndent> wtre unaccustomed tO giving demos.
`But they a~:reed to meer, on rhe porch of Choriton '• P..to Alto home,
`which Bechtohheirn p:used on his way 10 work each day.
`"David hac! a laptop on his porch in Palo Al<o, witb an Ethernet
`connectton," P.tge recalls. •we did a demo, and Andy asked a lot of
`questions. (Then) he said: 'WeU,l don't want to wane time. I'm sure
`it'll hdp you guy> if I ju>t write a cheek.' "
`Page and Brio """'en'< ready for such an offer, bU< when
`
`Bcchtolsheim wcnr our <o hi< = to get his checkbook. chey pon(cid:173)
`
`der«! how much to :uk for and ar wha< valuation. When Bechtol·
`shcim returned <hey mid him their suggested valuation. Page pick<
`up tilt SIOry: ·we told him our valuation, and he s:Ud 'Oh, I don't
`chink that's enough, I tbink it should be twice that much.' •
`Brin and Page were stunned, but of course, rhey ogreed, and
`Bechtolsheirn asked who the cheek should be made out to. The
`founders ha