in the probabilities can be computed in polynomial
`time.)
`
`z. = 1. Then. 1,-(1) = 1 + z. +z,. (Tu), 7, and .1, retain
`their original values).
`
`o 'I‘l1e value of a:‘,"'_ for other edges I may also be deter-
`mined. e.g., if :1” = I, then for all edges I adjacent
`to u. :7“ =
`
`Condition (4) holds since the changes in the probabilities at
`each step can be computed in polynomial time as mentioned
`- earlier. (Notice that the random variable I,- (I) is the sum of
`independent random variables). Condition (2) holds since
`
`It major stumbling block i11 applying the method of condi-
`tional probabilities is always the computation of the con-
`ditio11al probabilities.
`In our case, we do not compute the
`exact probability that there exists an overloaded edge (even
`initially). but rather only estimate it. Consequently, if the
`estimator is not chosen judiciously, it may happen that when
`a variable is considererl, according to the estimator, no value
`assigiied to it can lead to a good solution. To overcome
`this difficulty, following ltagliavaii [16], the notion of a pes-
`simistic estimator is introduced. We call P; a pessiiriisiic
`cslimalm'ol' the co11ditio11al probability
`if it satisfies the
`|’ollo\\-ing conditions:
`
`1. Po <1.
`
`2. For any partial assignment. of the first j variables,
`'3' 5 P1“
`
`3. min {P-,9,
`for i = 0. l.
`
`3 [554 where
`
`is the estimator ol’PJ-‘i
`
`4. The pessimistic esl.imators can be computed i11 poly-
`nomial time.
`
`lt is not very hard to see that such a pessimistic esti1na-
`tor can equally well he used iii the 1netl1od of conditional
`prol)abilil.ies instead of the exact conditional probabilities
`which are hard to compute i11 general. We now show that
`the pessimistic estimator that we will choose indeed satisfies
`the above conditions. We have earlier proved that initially,
`
`Prob [set is bad] 5
`
`)3 l’rob[I(f) > (1 + 1,171/)1
`IEE
`
`5
`
`E[¢M'(I)]
`————=—— < 1
`65 ,u+u1x,«</1
`
`Notice that )1, a11d 7, depend 011 the edge f. We deli11e
`
`Epmun
`P _
`° ' [GE ,u+u1Afl(I)
`
`The estimator at Step j is defined to be
`
`p.=
`’
`
`E[¢-‘I';'(l))
`______
`Z¢(l+7:)I(IW
`IEE
`
`where. I,-(f) is a random variable denoting the load on edge
`I at the end of Step j. For example, suppose that (U) =
`:r:. + 2;-_u + :3 + :4 and at the end of Step 1', :1 = 0 and
`
`P;
`
`s
`
`‘;ProI>1I,-(I)>u+n)7(n1
`[GE
`2 [§[¢4\/':‘(1)]-
`E ¢(l+‘I[)"l’(/)'
`
`= P-.
`J
`
`.;_°
`
`5
`
`Let us show that condition (3) holds as well. Suppose that
`at Step j+l variable z‘,“’ is being considered. By definition,
`
`Z E[¢*/'.'(!)]
`IE3
`
`pew . 2 E[5M';(/)|,,\:v = 1]
`[E15
`
`+ 11- P:") - 2 r:1eW"1z':" = 01
`Ice
`
`where the probability of choosing edge e as part of the path
`from u to v is P,“ (given the asignments of the previous j
`steps). Now,
`
`P:‘i+I = E
`1“:
`
`E[¢"Jli(/) lzgv = 1]
`¢(|+1/)7(l)*I
`E[¢M'ill)|,,-W = 0)
`,(1+m7(;)A1
`
`o
`_
`.
`P1“ ‘ 1:4;
`_
`.
`_
`Pi=PeW'Pji+1+(l‘P¢W)'Pjo+1
`
`.
`
`Hence,
`
`and clearly, min{P,§+,, P’-i+1)'5 Isj. The value of 2:" is set
`to the value for which l:“’5+,
`is minimized, for i = 0,1.
`
`4 Assigning Weights
`trolled Flooding
`
`for Can-
`
`ln this section we consider a more dynamic approach of
`routing—that of oontrolled flooding. Flooding is a routing
`strategy that guarantees fast arrivals with minimal enroute
`computation at the expense of excessive bandwidth use. To
`limit the extent of flooding we adopt the controlled flooding
`scheme first proposed in
`Considei a network in which
`each link is assigned a weight (sometimes referred to as car!)
`for traversing it and every message carries with it a wealth.
`A message arriving at an intermediate node will be dupli-
`cated and forwarded along all outgoing links (except the one
`it came from) whose cost is lower than the message wealth.
`The cost of the link is then deducted [mm the duplicated-
`message wealth. Consider for example the network in figure
`1 depicting a message with a wealth of 10 arriving at node 2.
`
`2A.4.6
`
`0175
`
`|PR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR
`Ex. 1102, p. 1322 of 1442
`
`i
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1322 of 1442
`
`

`
`
`
`Figure 1: Example of controlled flooding
`
`The link to node 3 has a cost oft‘) associated with it resulting
`in a copy of the message with wealth 4 to be transmitted
`along that link. Similarly. a copy of the message with a
`wealth of 0 will arrive at node 4. Nodes 5 and 6 will not
`receive a copy of the mewage.
`
`Since the controlled flooding scheme is a derivative of a
`flooding algorithm. it is impo§ihle to assure that a message
`always arrives only at the nodes it is intended to. In partic-
`ular. when used for point-to-point routing it is evident that
`more nodes than necessary might receive a message. In the
`above example, if the original message had arrived at node
`2 with a wealth of I3 node 4 would have received two copies.
`Note also that there is no way for node 1 to send a message
`to node 4 without node 3 also receiving it. Clearly. different
`weight assignments may change the pattern of flooding.
`
`The problem is to assign the link costs so as to achieve best
`performance. To that end a figure of Inerit is defined which
`is proportional to the (average) number of nodes that will
`'receive every mssage. An optimal weight assignment is one
`that minimizes the figure of merit. To formalize our discus-
`sion let the network he represented by the graph G(V. E)
`with |V| : n and |E| = m, let the length of a path in the
`network be defined as the sum of the weights of the edges
`of the path. and let. the shortest path between two nodes be
`the path with minimal length. Then, it is shown in [9] that
`for an assignment to be optimal, the following requirements
`(referred to as optimality requirements) must hold for every
`vertex (node) 1-:
`
`o For every vertex v E V, the shortest path from r to v
`is unique.
`
`0 For any two vertices u. v E V. the length of the short-
`est path from r to u is different from the length of the
`
`shortest path from r to u.
`
`Assignments that satisfy the above requirements are called
`good. An assignment is good with respect to r if allshortat
`paths from r satisfy the above requirements. Let us assume
`without loss of generality that the weights assigned are all
`positive integers. ‘
`
`Let |l . ..R] denote the range of numbers from which weights
`are drawn and let n denote the number of nodes in the net-
`work.
`If R = 2'5‘. it is easy to find a good assignment
`[9]. For example, assigning 2‘ as the weight of edge e; as-
`sures that any two different paths will have different lengths.
`However, because the length of the path is carried by every
`message it is desirable to reduce R as much as possible.
`
`We present two methods for constructing good assignments
`such that R is polynomial in n. In the first method the com-
`munication is restricted to a spanning tree T of the graph.
`This is done by assigning infinite weight to edges that are
`not in the tree. Denoting the tree edges by C], .
`. .e.- .
`.
`., the
`algorithm is recursively defined as follows. Let in be a leaf
`of T. let u; be its neighbor in the tree, and let eg be the edge
`connecting in and w.
`
`1. Compute (recursively) a good assignment for the tree
`T - U] .
`
`2. Extend the good assignment from T — II] to T.
`
`We assume inductively that a good assignment was com-
`puted in Step 1. Step 2 can be implemented by checking all
`the values in the range 1 .. .R and finding one that satisfies
`the requirements for a good assignment. Obviously, a good
`value for e; exists if R is large enough. The next lemma
`bounds the value of R.
`
`Lemma 4.1 If R 2 117,
`meut.
`
`than there exist: a good assign-
`
`Proof: Since a good assignment was computed for T — u.
`at Step 1, any value assigned to e; will complete a good
`assignment with respect to vi. The number ofdiatinct values
`that e; cannot assume is at most (1) — l)(n -— 2): for each‘
`vertex r E T—w. the distance from r to in should be different
`from the distance from r to any other vertex. and thus. there
`can be at most It — 2 forbidden values (with respect to r),
`and the claim follows.
`0
`
`The complexity of the weight afiignment algorithm is O(n°)
`since each step can be implemented in O(n') time. For each
`vertex 1/.‘ E V, a table of all its distances to the other vertices
`is maintained and for each node all the forbidden values
`in the range [1 ...n’] are marked. One of the unmarked
`numbers is chosen arbitrarily for-er. Then, the tables of all
`other noda are updated.
`
`0176
`
`2A.4.7
`
`n
`
`::..t .-
`
`
`
`|PR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex.1102, p. 1323 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1323 of 1442
`
`

`
`The above amignment, being tree based, makes no use of
`many of the networl: links. The second assignment, which
`we present next. has the property that the whole network
`participates in the communication. We prment two algo-
`rithms; the first
`is a«-randomized one that lends itself to
`distributed computation because the weight for each edge
`is chosen independently of the other edges. This algorithm
`generates a good assignment with high probability. The sec-
`ond algorithm is deterministic, and the weights are chosen
`from a smaller range than in the randomized algorithm.
`
`Our main tool in the randomized case is the Isolating Lemma
`ofMulmu|ey, Vazirani and Vazirani [10]. A set system (S. F)
`consists of a finite set S of elements, 5 = (:1. .. . .2"). and
`a family F of subsets of S, F = (S.,...,Sg). Let a weight
`in; be assigned to each element of S. The weight of a subset
`is defined to be the sum of the weights of its elements.
`
`Lemma 4.2 (Isolating Lemma) Lrl R Z n and let
`(S. F) be a set system whose _elcnienfs are assigned integer
`weights rlzarru uniformly and inrfepemfenfly from the range
`(l...Il]. Tlrrn, Prob[There is a unique minimum (maxi-
`mum) weight set in F] 2 I —
`
`(Note: the lemma in its original form in [10] was proven for
`R = 2n but actually holds for all R 2 1:).
`Cl
`
`We start by proving tlint the following randomized process
`will generate a good assignment with high probability. Let
`a weight for each edge be chosen randomly and uniformly
`from the range [1 .
`.
`
`Lemnu-\ 4.3 For It 3 11‘
`is good is al feast
`
`the probability that an assignment
`
`the shancst path between
`Let A.-,- be the event
`Proof:
`nodes 0; and v,-
`is not unique. Then A = U;_,‘/lg,‘
`is the
`event indicating the existence of at least one pair of nodes
`with non-unique shortest path between them. For each pair
`of nodes is; and 1-;
`let the set system F be the set of all
`paths connecting them. From the isolating lemma we have
`that the shortest path between them will be unique with
`probability at least
`1 —- 5,5, or, Pl'0l)(A.'j) 5
`llence,
`Prob[.-1) _<_ Z.-J. Prob[.-15,-] S
`-
`
`Let B.-,-. represent the event that nodes u.-, of. and 01,- form
`a bad triplet. namely that the length of the shortest path
`between rt; and 0;. equals that between I), and II),-. B =
`U.~,-1-B,-,1. then represents the existence of at least one bad
`triplet in the network. ln a way similar to the above we get
`Prob(B] 3
`-
`
`Finally. /I U D is the event indicating that the requirements
`are E met. and thus
`
`Prob[_r/oorf assignment]
`
`2
`
`I — Prob[A) — Prob[l3]
`
`Z
`
`_ n’(n — l)
`YR
`n'(n — l)(n — 2)
`6R
`
`.
`
`For R 3 n‘. the right handside exceeds
`
`D
`
`The last lemma providm us with a randomized distributed
`algorithm for constructing a good asignment. The proba-
`bility of failure can be made arbitrarily small by increasing
`the value of R.
`
`Notice that this method does not ensure that every edge
`participates in at least one shortat path. This can be fixed
`by forcing the weight assignment so that the BFS tree re-
`sulting from the weight assignment is also a EFS tree in
`the underlying graph without weights. To that end assign
`weights to the edges according to any of the above described
`algorithms and then add the value 11-}? to each weight. Now
`every edge takes part in at least one shortest path.
`'
`
`Next we show how a good assignment can be constructed
`deterministically. One way would be to derandomize the
`above randomized process. Notice that the proof of Lemma
`4.1 actually implies that every partial assignment that does
`not violate the optimality requirements can be completed
`to a good assignment. We can thus assign weights to the
`edgs one-by-one ensuring at every step that {tone of the
`requirements is violated.
`
`A better way of doing this is by the following algorithm that
`constructs a good assignment with R = n’ (compared with
`11‘). Initially, every edge e.- is assigned weight n‘ - 2‘. The
`weights of the edges are then changed one-by-one to fit into
`the range [1
`while maintaining the goodness of the
`assignment. At each step, the weight of the heaviest edge is
`changed.
`
`Lemma 4.4 If R 2 n3,
`structed.
`
`:1 good assignment can be con-
`‘
`
`Proof: The invariant which is maintained at the end of
`each step is that the assignment remains good. This is true
`initially. Let we be the new weight assigned to edge er at
`step i, where e.- connects vertices 1: and y. We prove that
`to; can be fitted into the range (1.. .R] by bounding the
`number of forbidden values for w.- and showing that at least
`one permitted number exists. Let f... denote the value of
`the shortest distance between vcrtex..u. and vertex u when
`edge e; is removed from the graph (1.... might be infinite).
`
`To maintain goodness we must accommodate both optimal-
`ity requirement. We first show how to maintain the unique-
`ness of the shortest path between every pair of vertica. Let
`r and u be a pair of vertices, and assume without loss of
`generality that 1,, < I..,,.
`(They cannot be equal by the
`invariant). If the removal of edge c.- from the graph leaves
`
`2A.4.8
`
`0177
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR
`Ex. 1102, p. 1324 of 1442
`
`i
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1324 of 1442
`
`

`
`vertica r and u in different connected components. then any
`value can be chosen for w; with rapect to r and v. Assume
`this is not the case. Since edge e.- had the largat weight
`in the graph (i.e.. n‘ - 2‘), the shortest path from r to 1:
`cannot contain edge c.- and l,.,
`is the value of the shortest
`distance from r to 11.. Hence. to maintain the _uniqueness of
`the shortest path requirement. it is enough that
`
`Irv # lrx 'l" wi + lync-
`
`(Noticc that the shortest path will remain unique even if it
`contains edge e.-, because of the uniqueness of the shortest
`paths from r to 2 and from y to 0). This condition generates
`at most in - 1 forbidden values for ID; with respect to every
`vertex 1' in the graph, or n(n- l) forbidden values altogether.
`
`Let us now show how the second requirement of optimality
`is maintained. Let r. u and ll be a triplet of vertices. Again,
`notice that if the removal of edge 2; from the graph leaves
`vertex r in one connected component. and vertices u and
`u in a different connected component. then any value can
`be chosen for In; with respect
`to r, u and v. The same
`holds if the removal of e.- leaves y separated from r, u. and
`1:. Assume this is not the case.
`lt follows from the above
`discussion that the shortst distance from r to u is either
`1,... or I.., + uv.~ +1’... Similarly. the shortest distance from
`r to u is either 1,... or In + w.- + l,.,.
`
`By the invariant.
`
`l,., + m.- +1," ;6 In + w; + 1,...
`and
`I..., # l,.,
`Hence. to maintain the second requirement of optimality, it
`is enough that
`
`and
`
`lrv ¢ Ir: + WI" + lyu
`
`lru ¢ lrx 'l" Wu‘ + ‘yu-
`
`These two conditions add at most 2 - (";') forbidden values
`for III,‘ with respect to every vertex 1' in the graph, for a total
`of‘2n . ('-;*).
`
`Altogether. the number of forbidden values for ‘w;
`l)(n + l) < via. and the lemma follows.
`
`is n(n —
`0
`
`Note that the initial assignment (e.- = ti‘ ~2‘) is chosen to
`ensure that every edge is treated exactly once, and when it
`is treated it does not participate in any shortest path unless
`it is a bridge.
`
`The complexity of the algorithm is O(u3m) since each step
`can be implemented in 0015) time. Every vertex v.- E V
`maintains a table with all its shortest distances to the other
`vertices; it then marks all the forbidden values in the range
`[I .. .113]. One of the unmarked numbers is chosen arbitrarily
`for c.-. Then, the. tables of all other vertices are updated.
`
`The reason why the range can be made smaller in the deter-
`ministic case is that it is enough to ensure at each step that
`
`there is one good value, whereas in the randomized case.
`one has toensure success with high probability.
`
`A desirable property of a routing scheme is having the traffic
`be evenly distributed among the edga. Unfortunately, this
`is the drawback of routing with random weights. The follow-
`ing example shows that with high probability this scheme
`does not yield a balanced load.
`
`Let the load on an edge be defined as the number of short-
`est paths that contain it, and consider a graph made of
`two cliques of size k that are interconnected by two edges,
`e. and e;. The weight for each edge is chosen uniformly
`and independently from the range [1 . . .R].
`ln each clique.
`the distribution of the weights is uniform and thus.
`if the
`weights of e; and C2 are not close to one another, most of
`the traffic between the two cliques would go through the
`edge with smaller weight. Since this event will happen with
`high probability, the communication would not be balanced
`with high probability.
`
`5 Conclusion
`
`In this paper we examined several routing strategies for fast
`modern packet switching networks. The relevant charac-
`teristic of these networks is the inability to make elaborate
`routing decisions while packets are being switched. At the
`switching speeds being considered, looking up a table whose
`size is proportional to the number of network nodes is con-
`sidered too costly.
`
`These requirements limit the number of applicable routing
`strategies. The simplest and most natural strategy is to
`use fixed routing schemes in which the route between every
`pair of source-destination nodes is fixed in advance. The
`problem would then be to find a set of routes so that net-
`work resources are utilized as evenly as possible. Two such
`strategies are analysed in this paper:
`routing along trees
`and routing along paths. For both cases polynomial algo-
`rithms are devised. we show that in both cases no network
`link remains unused but that routing along paths is likely
`to be a better strategy from load balancing standpoint.
`
`Deviating from the fixed routing sdieme we analyze a con-
`trolled flooding scheme in which every message essentially
`floods the networks but the extent of its flooding can be
`controlled by link weights. We provide a polynomial algo-
`rithm to compute thse weights but show that the scheme
`cannot guarantee a good balance of load.
`
`0178
`
`2A.4.9
`
`|PR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1325 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1325 of 1442
`
`

`
`Acknowledgement
`
`[13] M. Carey and D. Johnson, Computers and Intructa6il-
`ity. San Francisco: W.ll. Freeman and Company, 1979.
`
`W- would like to l.h:\nk Noga Alon for many helpful discus-
`sions on this paper and in particular for his help in analyzing
`the algorithm of Section 2.1.
`
`[14] D. Angluin and,L. G. Valiant, “Fast probabilistic algo
`rithms for hamiltonian circuits and matching,‘ Jour-
`nal of Computer and System Sciences, vol. 18, pp. 155-
`l93, 1979.
`
`[15] 1. Spencer, Ten Lectures on the Probabilistic Method.
`Philadelphia, Pennsylvania: SIAM, I987.
`
`[16] P. Raghavan, “Probabilistic construction of determin-
`istic algorithms: Approximating packing intcger pro
`grams,” Journal of Computer and System Sciences,
`vol. 37, pp. l30-l43,0ctobcr 1988.
`
`References
`
`[l] A. F.phtcmidcs, "Flie routing problem in computer net-
`works," in Cannnunications and Networks (l. Blake and
`II. Poor, cds.). pp. 299-324, New York: Springer Ver-
`lng, 1986.
`
`[2] M. Schwartz and '1‘. Stern, “flouting techniques used
`in computer communication networks," IEEE Trans.
`on C-'arnmunimtions, \'ol..COl\‘l-28, pp. 539-555, April
`I980.
`
`[3]
`
`[4]
`
`I’. Green, “Computer communications: Milestones and
`prophecies." IEEE Communications, pp. 49-63, 1984.
`
`l. Cidon and l. Gopnl, “Paris: An approach to in-
`tegratcd high-speed private Iietworks," lirtenratiarial
`Journal of Digital and Analog Cabled Systems, vol. 1,
`pp. 77-86. April-Jun:-. W88.
`
`[5] J. Turner, “Design ofa broadcast. packet switching net-
`work." IEEE Trans. on Comuumicatious, vol. COM-
`-'lG. pp. 734-743. June 1988.
`
`[(5]
`
`Inlercomiectian Networks for Lnrye-Scale
`ll. Sicgcl,
`Pamllcl Processing: Theory and Case Studies. Lex-
`ington. MA: Lexington Books. 1984.
`
`[T] L. l"ral.ta, M. Gcrla, and L. l\'|cinrock, “The flow devi-
`ation method: An approach to store and forward com-
`munication network design," NetmorL'.s, vol. 3, no. 2,
`pp. 97-133, I973.
`
`[8] R. Gallager. “A minimum delay routing a.lgoritlnn us-
`ing distributed computation," IEEE Trans. on Cam-
`muuimtious, vol. COM-25, pp. 73-85, January 1977.
`
`[9] O. Lesser and R. Rom, “Routing by controlled flooding
`in communication networks," in Proceedings of IEEE
`In/ncnm '90, (San Francisco, California), pp. 910-917,
`IEEE, June 1990.
`
`[I0]
`
`l\'. Mnlmnlcy. U. Vazirani, and V. Vazirani, “Matching
`is as easy as iiiatrix inversion,” Combinatorica, vol. 7,
`no.
`I‘, pp. 105-H3. 1987.
`
`[ll] J. llopcroft and It. harp, “An :15” algorithm for max-
`imum nmtching in bipartite graphs,” Siam J. Comput-
`ing, vol. 2, pp. 225-23], 1973.
`
`[I2] S. Even, Graph Algorithuu. New York: Computer Sci-
`ence Press, 1979.
`
`2A.4.l0
`
`0179
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR
`Ex. 1102, p. 1326 of 1442
`
`i
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1326 of 1442
`
`

`
`0-«
`
`'-
`
`‘Q "nu
`
`imp://proquest.umi.convpqdweb7rs=1os2s...2&nrp=1&Did=oooooo14o2s2s3 l&Mtd=l &Fm
`
`e
`Bus(;vsW?re
`
`Re Global Leader in :\"eu.': l)i-rlribulitm
`
`Boeing and Panthesis Complete SWAN Transaction
`Business Wire; New York; Jul 22, 2002; Business Editors & Aerospace Writers;
`
`NAICS:336411 NAICS:3364l3 NAICS:336414 Duns:00-92_5-6819
`Start Page:
`1
`Companies: Boeing Co Ticker:BA Duns:00-925-6819 NAICS:336411 NAICS.'336413
`NAICS:336414
`
`Abstract:
`
`IRVINE, Ca/if.—-(BUSINESS WIRE)--July 22, 2002--The Boeing Co. and Panthesis Inc., today
`announced that they have completed a transaction that gives Boeing an equity stake in Panthesis
`and provides Panthesis with an exclusive right to commercialize Boeing's Small- world Wide Area
`Networking (SWAN) technology.
`
`Based in Bellevue, Wash., Panthesis, was established in 2001 to develop and commercialize
`innovative software technology. Its co- founders, current Chief Development Officer Dr. Fred Holt
`and Chief Technology Officer Virgil Bourassa, are both former employees of The Boeing Co., where
`they co-invented SWAN technology while working in the Mathematics and Computing Technology
`unit of the Boeing Phantom Works R&D division.
`'
`
`Full Text:
`
`Copyright Business Wire Jul 22, 2002
`
`IRVINE, Calif.--(BUSINESS WIRE)--July 22, 2002--The Boeing Co. and Panthesis Inc., today
`announced that they have completed a transaction that gives Boeing an equity stake in Panthesis and
`provides Panthesis with an exclusive right to commercialize Boeing's Small-world Wide Area
`Networking (SWAN) technology.
`
`SWAN technology was originally developed by Boeing to allow multiple geographically dispersed
`people to conduct collaborative meetings and engineering design reviews in real time. ‘
`‘
`
`"SWAN is a revolutionary technology that can be used to enhance numerous computing, networking and
`communications functions," said Linda Magnotti, CEO of Panthesis. "The sophisticated mathematics and
`software architecture underlying SWAN technology can provide reliable server—less communication for
`communities anywhere in the world."
`
`Magnotti added that Panthesis is currently focusing its development efforts on providing the bandwidth
`multiplication needed for use in massive multi-player online games, real-time online auctions, content
`distribution and other large—scale, unlimited online collaborations.
`
`Based in Bellevue, Wash., Panthesis, was established in 2001 to develop and commercialize innovative
`software technology. Its co- founders, current Chief Development Officer Dr. Fred Holt and Chief
`Technology Officer Virgil Bourassa, are both former employees of The Boeing Co., where they
`co-invented SWAN technology while working in the Mathematics and Computing Technology unit of
`the Boeing Phantom Works R&D division.
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1327 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1327 of 1442
`
`

`
`Dgcumem
`
`http://proquest.umi.com/pqdweb?TS= 1 0528...2&Dtp= l &Did=0000OO14028253 I &Mtd=1 &Fm
`
`"Because Panthesis clearly has the expertise for adapting SWAN technology to a broad range of potential
`applications, we were confident in giving them the exclusive right to commercialize this technology in
`the global marketplace," explained Gene Partlow, vice president of Boeing's Intellectual Property
`Business.
`
`The potential for this agreement was created through Boeing's Chairman's Innovation Initiative, which
`promotes the development of new business ventures based on entrepreneurial ideas from employees.
`While some ideas are developed into spin-off companies, others are spun into Boeing business units for
`further development or, like SWAN, into the Intellectual Property Business for other types of business
`transactions.
`
`Panthesis is currently seeking investment capital to support company expansion and market penetration,
`and is engaged in developing relationships with key customers in the online auction and gaming markets.
`
`The Boeing Co., with headquarters in Chicago, is the world's leading aerospace company and the No. 1
`U.S. exporter. It is the largest manufacturer of satellites, commercial jetliners and military aircraft, and it
`provides a full range of lifecycle support for these and other products. The company is also a global
`market leader in missile defense, human space flight and launch services. Boeing capabilities also
`include financial services and advanced information and communications systems.
`
`
`
`Reproduced with permission of the copyright owner. Further reproduction or distribution is prohibited without permission.
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1328 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1328 of 1442
`
`

`
`-‘.!t_)cument
`
`\‘-
`
`http://proquesLumi.com’pqdweb'!TS= l 05 28...2&Dtp= l &Did=00000002899 l 200&Mtd= l &.Fm
`
`-2.
`
`Microsoft Boosts Accessibility to Internet Gaming Zone With Latest Release
`PR Newswire; New York; Apr 27, 1998;
`
`1
`Start Page:
`Dateline: Washington
`Companies: Microsoft Corp
`Abstract:
`
`REDMOND, Wash., April 27 /PRNewswire/ -- Microsoft Corp. (Nasdaq: MSI-T) today released its
`latest update for the Microsoft(R) Internet Gaming Zone ( http://www.zone.com/ ), featuring
`support for Netscape 4.0 and the latest versions of Microsoft Internet Explorer. The new version
`makes the Zone accessible to the majority of Internet users. With this new version, the Zone also
`introduced the new Zone Rating System, which allows game players to determine how they fare
`against other players. Chess and Age of Empires(R) will be the first games with the Zone Rating
`System, and new games are scheduled to be added to the system in the coming weeks.
`
`The Zone is a collective place for gamers to play today's best games against others for free. Players
`have a wide variety of games to choose from -- including parlor games like Hearts and Chess, and
`action and strategy games like Jedi Knight: Dark Forces II, Age of Empires and the Fighter Ace(TM)
`online multiplayer game, the site's first premium game designed specifically for massive multiplayer
`gaming via the Internet. Furthermore, visitors can navigate through the site before downloading
`the Zone software required for game play.
`
`Full Text:
`
`Copyright PR Newswire - NY Apr 27, 1998
`
`Industry: COMPUTER/ELECTRONICS; INTERNET MULTHVIEDIA ONLINE
`
`Netscape Support and Player Rating System Featured in Newest Version
`
`Of the Leading Internet Gaming Site
`
`REDMOND, Wash., April 27 /PRNewswire/ -- Microsoft Corp. (Nasdaq: MSFT) today released its
`latest update for the Microsofi(R) Internet Gaming Zone ( http:llwww.zone.con1/ ), featuring support for
`Netscape 4.0 and the latest versions of Microsoft Internet Explorer. The new version makes the Zone
`accessible to the majority of Internet users. With this new version, the Zone also introduced the new
`Zone Rating System, which allows game players to determine how they fare against other players. Chess
`and Age of Empires(R) will be the first games with the Zone Rating System, and new games are
`scheduled to be added to the system in the coming weeks.
`
`"We believe online gaming is all about social interaction with a large and active community," said Ed
`Fries, general manager of the games group at Microsoft. "So we're very pleased that this new version of
`the Zone provides access for virtually everyone online."
`
`Already home to nearly 1.5 million online gamers, the Zone has more than 7,500 simultaneous users at
`peak times -- and is gaining new registered members at the rate of one every 20 seconds.
`
`The Zone is a collective place for gamers to play today's best games against others for free. Players have
`a wide variety of games to choose from -- including parlor games like Hearts and Chess, and action and
`strategy games like Jedi Knight: Dark Forces II, Age of Empires and the Fighter Ace(TM) online
`multiplayer game, the site's first premium game designed specifically for massive multiplayer gaming
`via the Internet. Furthennore, visitors can navigate through the site before downloading the Zone
`software required for game play.
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1329 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1329 of 1442
`
`

`
`oscumem _
`
`,
`,_~' "
`
`http://proquest.umi.com/pqdweb?TS=l0528...2&Dtp=l&Did=00000002899l200&Mtd=l&Fm
`
`In addition to Netscape 4.0 support and the Zone Rating System, the newest version of the Zone also
`features a new, streamlined interface, which reduces download times and makes getting into a game
`even easier. The Zone further assists its members with improved help and chat features.
`
`Variety and Popularity of Games Drive Growth
`
`The Zone offers a popular variety of classic card and board games such as Spades, Bridge and
`Backgammon. In fact, Spades has grown to become the most popular game on the Zone with peak usage
`of more than 2,000 players. In the past year, the Zone's lineup of CD-ROM games with free
`matchmaking has expanded rapidly with the addition of such popular Microsoft games as Age of
`Empires and Flight Simulator 98, and other top titles such as Jedi Knight: Dark Forces H from LucasArts
`Entertainment Co., Quake H from id Software and Scrabble from Hasbro Interactive, a unit of Hasbro
`Inc. These additions have brought the total number of games available for play on the Zone to 32. The
`Zone also recently announced support for upcoming Tom Clancy titles Rainbow Six and Dominant
`Species from Red Storm Entertainment.
`
`The Internet Gaming Zone has served Internet gamers since October 1995. In May 1996, Microsoft
`acquired Electric Gravity Inc., the original designer of the Internet Gaming Zone. The Internet Gaming
`Zone offers free membership with three components: free classic card and board games, free
`matchmaking for retail games, and access to premium games designed exclusively for the Zone
`(connect-time charges may apply). Most recently, Microsofi launched Fighter Ace, a World War H aerial
`combat premium game designed specifically for the Internet in which more than 100 players can
`dogfight in a single flight arena.
`
`Founded in 1975, Microsofi is the worldwide leader in sofiware for personal computers. The company
`offers a wide range of products and services for business and personal use, each designed with the
`mission of making it easier and more enjoyable for people to take advantage of the full power of
`personal computing every day.
`
`For online product infonnation:
`
`Microsoft Web site: http://www.microsofl.com/
`
`Microsoft Internet Gaming Zone Web site: http://www.zone.com/
`
`NOTE: Microsofi, Age of Empires and Fighter Ace are either registered trademarks or trademarks of
`Microsoft Corp. in the United States and/or other countries. Other product and company names herein
`may be trademarks of their respective owners. SOURCE Microsoft Corp.
`
`
`
`Reproduced with pemiission of the copyright owner. Further reproduction or distribution is prohibited without permission.
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1330 of 1442
`
`IPR2016-00726 -ACTIVISION, EA, TAKE-TWO, 2K, ROCKSTAR,
`Ex. 1102, p. 1330 o