MICHEL, Circuit Judge.

Bell Communications Research, Inc. ("Bellcore") appeals from the August 16, 1994 decision of the United States District Court for the District of New Jersey, No. 9S 4104, granting summary judgment in favor of Vitalink Communications Corporation ("Vitalink") on the latter's counterclaim for a declaration of noninfringement and dismissing Bellcore's infringement suit against Vita link. Because the trial court partially misconstrued the scope of the asserted claim of U.S. Patent No. 4,706,080 ('080), and consequently erred in granting summary judgment of noninfringement, we vacate and remand.

BACKGROUND

A. The Technology

Local Area Networks ("LANs") consist of a number of devices, such as computers or telephones, attached to a shared communications medium. The communications medium permits the devices to transmit bundles of data, or "packets," back and forth to one another. Such packets, which contain fields of information that function as source device and destination device addresses, are broadcast through the communications medium; only the device that recognizes its own address as the destination address receives the broadcast.

The quality of a LAN's performance degrades in proportion to both the number of devices in the network and the speed at which each device processes information. One can, however, recover some of this lost performance by creating networks of multiple LANs rather than simply adding devices to a single, larger LAN. These multiple LANs are connected by means of "bridges," each of which are themselves composed of two paired "gateways." The gateways, each of which have memory capacity, maintain running lists of the source addresses of the packets they have forwarded, allowing the bridges they form gradually to "learn" how to broadcast packets selectively in order to reduce the network's overall load.

The existence of multiple pathways between a given pair of devices in two different LANs creates the potential for loops and thus thwarts the learning function of memory-capable gateways — a packet that cycles through a complete loop causes its source address to appear on both sides of the relevant gateway pairs, eliminating the advantage of equipping gateways to keep lists of source addresses from the packets they have broadcast. For purposes of describing and solving this looping problem, one can depict a group of interconnected networks as a graph, with lines and vertices used to represent connecting bridges and connected LANs, respectively. In this graphical notation, a "tree" is a graph in which a sequence of one or more lines connects two vertices, while a "spanning tree" is a graph in which all the vertices are connected. Such a spanning tree can be superimposed on the complete graph of the bridged networks and used to determine a set of loop-free spanning tree paths among the vertices. This process of determination can be accomplished either by some oversight mechanism or automatically by the bridges.

While the use of one spanning tree deals with multiple path and looping problems, two problems remain: spanning tree backup paths remain inactive unless bridge failures require that they be used, and the spanning tree's root may become a performance bottleneck for the system. The use of multiple spanning trees thus represents an improvement over the use of only one spanning tree. But one cannot implement a system for the use of multiple spanning trees without some means of differentiating among the spanning trees, such that a packet is always forwarded over some tree. A system can achieve the required differentiation in one of a number of ways: for example, the system could randomly assign different device addresses to different trees, or, alternatively, the source device could specify a tree by means of a tree identifier when it originates the packet.

B. The Patent in Suit

Bellcore's '080 patent, entitled "Interconnection of Broadcast Networks," discloses a method for interconnecting networks, such as LANs, that uses multiple concurrent spanning trees for packet delivery while preserving loop-free paths. According to the summary of the invention contained in the specification,

[e]ach spanning tree is uniquely identified. Each message packet that traverses the overall system is assigned to a specific spanning tree so the packet travels between nodes [ i.e., device networks] along edges [ i.e., bridges] contained in the specified spanning tree. Each gateway, with an expanded store-and-forward protocol [in its memory], parses the packet to determine the assigned spanning tree and forwards the message accordingly. In one embodiment of the present invention, the device originating the packet specifies the spanning tree identifier and conveys it either explicitly or implicitly in the packet.

Col. 2, 11. 14-25. As the more detailed description explains,

To implement the improvement in the gateway protocol arrangement in accordance with one aspect of the present invention, a set of spanning trees is selected for the cyclic graph according to predetermined guidelines. Each spanning tree is assigned a unique identifier or number and each message traversing the system is assigned to a unique spanning tree via its identifier. Any gateway receiving this message determines the tree number and then routes the message over the specified spanning tree and drops all packets of other spanning trees. Typically, the device originating the message specifies the spanning tree number, either explicitly or implicitly. For instance, with the explicit approach, a "tree number" field could be added to the packet specifications, say as an extra bit in the header of the packet. With the implicit approach, a spanning tree number could be generated from fields normally occurring in the packet such as the source and destination addresses. An appropriate example function might be

spanning tree number = (source `exclusive or' destination) modulo N,

where N is the number of spanning trees in the network. This has the benefit that all traffic between a pair of hosts will travel on only one spanning tree, thus minimizing the occupied drop lists across the system.

Col. 5, ll. 11-35. Bellcore's method also preserves the "transparency" of the interconnections among the LANs, according to which the existence of gateways does not require modifications to the networked devices or the packets they broadcast. Col. 1, ll. 29-35.

Claim 6 of the '080 patent, the only claim asserted by Bellcore, reads as follows:

6. A method for transmitting a packet over a system comprising a plurality of networks interconnected by gateways, said packet originated by a source device connected to one of said networks and destined for a destination device connected to one of said networks, said packet including a source address and a destination address, and said method comprising the steps of

defining an undirected graph representative of the system wherein said networks comprise graph nodes and said gateway[s] comprise graph paths,

defining a spanning tree on said graph such that every pair of said nodes is connected by only one of said paths and selecting a plurality of spanning trees for said graph according to pre-determined system guidelines,

configuring each gateway with source address lists in correspondence to the number of trees having said each gateway comprising one of said paths, wherein said lists reduce to a common list whenever said selection of spanning trees results in identical ones of said lists for said each gateway,

assigning, by said source device, one of said trees to broadcast said packet and associating with said packet an identifier indicative of said one of said trees,

broadcasting said packet by said source device through the system on said one of said trees, and

for each gateway receiving said packet,

(i) determining for each said packet said source address, said destination address and said packet identifier,

(ii) if said receiving gateway does not process packets having said identifier, inhibiting forwarding of said packet; otherwise, inserting said source address in the corresponding one of said lists associated with said identifier, and

(iii) inhibiting forwarding of said packet if said destination address is in said corresponding list; otherwise, for warding said packet by said receiving gateway.

Col. 10, ll. 18-57.

C. The Accused Products

Vitalink manufactures and markets communications products, including bridges and bridge-routers used in the networking of LANs. Specifically, Vitalink sells a series of products including a Distributed Load Sharing ("DLS") feature, which is itself the subject of a Vitalink patent. These products are alleged by Bellcore to use the method claimed in the '080 patent.

Both Bellcore and Vitalink agree that, in a Vitalink product using the DLS feature, a message packet contains both a source address and a destination address but no separate packet identifier. In other words, DLS uses only the implicit approach to spanning tree identification in the packet. The parties also agree that the tree along which a message packet travels in a DLS system may change mid-course in response to such phenomena as link failures and the opening of additional links or, when the system is in a steady state, remain the same throughout its transmission from the source device to the destination device.

D. The Infringement Suit

Bellcore originally sued Vitalink in the Eastern District of Virginia, alleging, among other things, that Vitalink's DLS products infringe Claim 6 of the '080 patent, either literally or under the doctrine of equivalents. Vitalink counterclaimed for a declaration of, among other things, noninfringement. After the case was transferred from Virginia to New Jersey on Vitalink's motion, the parties exchanged extensive written discovery, including affidavits from opposing computer science experts. After the close of discovery, Vitalink moved for summary judgment on its counterclaim for a declaration of noninfringement, contending that "in the DLS system a message packet is not assigned to a specific spanning tree by a source device as required by claim 6 . . . [and] a spanning tree identifier is not associated with the message packet as required by claim 6."

The district court, in a letter opinion dated August 16, 1994, granted Vitalink's motion for summary judgment of noninfringement and dismissed Bellcore's complaint. The court's decision in Vitalink's favor followed from its construction of Claim 6, according to which "the language of Claim 6 reveals that the Claim 6 packet literally requires three separate elements: (1) a source address, (2) a destination address, and (3) a packet identifier." Slip op. at 9. In addition, the district court construed Claim 6 to require that the source device assign, prior to transmission, the one spanning tree along which the packet travels to its specified destination. Id. at 8. Having construed Claim 6 to require both the assignment of one and only one spanning tree route prior to broadcast and the insertion into the packet of a spanning tree identifier separate from the destination and source addresses, the district court concluded that Bellcore could not make out a case of literal infringement against Vitalink. According to the district court, "the accused Vitalink DLS products, while having a source address and a destination address, do not employ a packet identifier contained within its packet such that the all elements or all limitations rule of literal infringement cannot be satisfied in the present case." Id. at 10-11. Similarly, with respect to the assignment step of Claim 6, the court noted that "Bellcore has effectively conceded that after the message packet has been broadcast by [a Vitalink] DLS source device, events might well intervene to alter the packet's route in mid-course . . . and hence the source device has no ability to assign the spanning tree along which the message packet is to be routed," id. at 11-12, again precluding literal infringement. Finally, the district court concluded that Bellcore's claim of infringement under the doctrine of equivalents was legally deficient.

Bellcore appeals from the district court's decision, contending that Claim 6, when properly construed, covers both the explicit and implicit approaches to spanning tree identification in the packet, as described in the specification, and neither provides for nor rules out the possibility of mid-course alterations in a packet's path within a system embodying the claimed method.

DISCUSSION

The moving party is entitled to summary judgment under Federal Rule of Civil Procedure 56(c) "if the pleadings, depositions, answers to interrogatories, and admissions on file, together with the affidavits, if any, show that there is no genuine issue as to any material fact and that the moving party is entitled to a judgment as a matter of law." FED.R.CIV.P. 56(c). We review the district court's grant of summary judgment de novo, Conroy v. Reebok Int'l, Ltd., 14 F.3d 1570, 1575, 29 USPQ2d 1373, 1377 (Fed. Cir. 1994), resolving all doubt respecting the presence or absence of genuine factual issues in the nonmovant's favor, Union Carbide Corp. v. American Can Co., 724 F.2d 1567, 1571, 220 USPQ 584, 588 (Fed. Cir. 1984). Finally, the disposition of this appeal turns largely on whether the trial court properly construed Claim 6 of the '080 patent, a question of law that we review de novo. Markman v. Westview Instruments, Inc., 52 F.3d 967, 978-80, 34 USPQ2d 1321, 1329 (Fed. Cir. 1995) (in banc).

A. Claim Construction

Before turning to the parties' contentions about the proper construction of the asserted claim, it is important to review some basic principles of claim construction. First, and most importantly, the language of the claim defines the scope of the protected invention. Yale Lock Mfg. Co. v. Greenleaf, 117 U.S. 554, 559, 6 S.Ct. 846, 847, 29 L.Ed. 952 (1886) ("The scope of letters-patent must be limited to the invention covered by the claim, and while the claim may be illustrated it cannot be enlarged by language used in other parts of the specification."); Autogiro Co. of Am. v. United States, 384 F.2d 391, 396, 155 USPQ 697, 701 (Ct.Cl. 1967) ("Courts can neither broaden nor narrow the claims to give the patentee something different than what he set forth [in the claim]."). See also Continental Paper Bag Co. v. Eastern Paper Bag Co., 210 U.S. 405, 419, 28 S.Ct. 748, 751, 52 L.Ed. 1122 (1908); Cimiotti Unhairing Co. v. American Fur Ref. Co., 198 U.S. 399, 410, 25 S.Ct. 697, 702, 49 L.Ed. 1100 (1905). Accordingly, "resort must be had in the first instance to the words of the claim," words to which we ascribe their ordinary meaning unless it appears the inventor used them otherwise. Envirotech Corp. v. Al George, Inc., 730 F.2d 753, 759, 221 USPQ 473, 477 (Fed. Cir. 1984). Second, it is equally "fundamental that claims are to be construed in the light of the specifications and both are to be read with a view to ascertaining the invention." United States v. Adams, 383 U.S. 39, 49, 86 S.Ct. 708, 713, 15 L.Ed.2d 572 (1966). See also Markman, 52 F.3d at 978-80, 34 USPQ2d at 1329-30 ("Claims must be read in view of the specification, of which they are a part.... For claim construction purposes, the [specification's] description may act as a sort of dictionary, which explains the invention and may define terms used in the claims.").

We construe claim preambles, like all other claim language, consistently with these principles. Much ink has, of course, been consumed in debates regarding when and to what extent claim preambles limit the scope of the claims in which they appear. See, e.g., 2 DONALD S. CHISUM, PATENTS § 8.06[1][d] (1993); 1 ANTHONY W. DELLER, PATENT CLAIMS §§ 78, 163-83 (2d ed. 1971); Willis Higgins, The Significance of Preambles in Chemical Composition Claims, 49 J.PAT. TRADEMARK OFF. SOC'Y 337 (1967); Vincent Millin, PTO Practice: Preamble — Prelude to Patentability, 72 J.PAT. TRADEMARK OFF. Soc'Y 348 (1990); David R. Pressman, Note, Patents — Claim Construction, 30 GEO.WASH.L.REV. 380 (1961); Alton D. Rollins, Is It New or Not.?, 68 J.PAT. TRADEMARK OFF. Soc'Y 89 (1986). These debates center, however, on particular arts and claiming styles and do not call into doubt the general principle, as well-settled as any in our patent law precedent, that a claim preamble has the import that the claim as a whole suggests for it. In other words, when the claim drafter chooses to use both the preamble and the body to define the subject matter of the claimed invention, the invention so defined, and not some other, is the one the patent protects. In re Paulsen, 30 F.3d 1475, 1479, 31 USPQ2d 1671, 1673 (Fed. Cir. 1994) ("[T]erms appearing in a preamble may be deemed limitations of a claim when they give meaning to the claim and properly define the invention.") (internal quotation omitted); London v. Carson Pirie Scott Co., 946 F.2d 1534, 1539, 20 USPQ2d 1456, 1459 (Fed. Cir. 1991) ("The shank is defined in the preamble as that portion of the hanger `between the supporting hook for the hanger and the support for the garment.' This is not merely a suggested use or `clarifying language,' as London argues, but rather a limitation supported by structure which must be satisfied by Samsonite's clamp, either literally or equivalently[,] if infringement is to be found."); In re Stencel, 828 F.2d 751, 754, 4 USPQ2d 1071, 1073 (Fed. Cir. 1987) ("Whether a preamble of intended purpose constitutes a limitation to the claim is, as has long been established, a matter to be determined on the facts of each case in view of the claimed invention as a whole."); Loctite Corp. v. Ultraseal Ltd., 781 F.2d 861, 866, 228 USPQ 90, 92 (Fed. Cir. 1985) ("Although it appears in the preambles of the '012 patent claims, the term `anaerobic' breathes life and meaning into the claims and, hence, is a necessary limitation to them."); Perkin-Elmer Corp. v. Computervision Corp., 732 F.2d 888, 896, 221 USPQ 669, 675 (Fed. Cir.) ("The system of claim 1 is one of unity magnification and is image forming. Those limitations appear in the preamble, but are necessary to give meaning to the claim and properly define the invention."), cert. denied, 469 U.S. 857, 105 S.Ct. 187, 83 L.Ed.2d 120 (1984). One of our predecessor courts summarized this approach in Kropa v. Robie, 187 F.2d 150, 88 USPQ 478 (CCPA 1951), after reviewing some 37 cases that turned on the limiting nature of the preambles to the claims in suit. According to the court in Kropa,

the preamble has been denied the effect of a limitation where . . . the claim or [interference] count apart from the introductory clause completely defined the subject matter [of the invention], and the preamble merely stated a purpose or intended use of that subject matter. On the other hand, in those . . . cases where the preamble to the claim or count was expressly or by necessary implication given the effect of a limitation, the introductory phrase was deemed essential to point out the invention defined by the claim or count. In the latter class of cases, the preamble was considered necessary to give life, meaning and vitality to the claims or counts.

Id. at 152, 88 USPQ at 480-81. Preamble construction thus presents no deeper mystery than the broader task of claim construction, of which it is but a part.

B. Claim 6 of the '080 Patent

Claim 6 of the '080 patent recites a "method for transmitting a packet over a system comprising a plurality of networks ... said packet including a source address and destination address," as its preamble indicates. It then recites, inter alia, the steps of "assigning, by said source device, one of said trees to broadcast said packet and associating with said packet an identifier indicative of said one of said trees." (Emphasis added). These two steps of the claimed method, by referring to " said packet," expressly incorporate by reference the preamble phrase "said packet including a source address and a destination address." As a result, only a method for transmitting packets that have both source and destination addresses can literally infringe Claim 6.

Bellcore contends, as it did before the district court, that one ought not to accord "definitional status" to the phrase "said packet including a source address and a destination address" because it appears in the claim's preamble, relying primarily on our decision in DeGeorge v. Bernier, 768 F.2d 1318, 226 USPQ 758 (Fed. Cir. 1985), for the proposition that "[t]he preamble to a claim does not limit the claim." Bellcore likely maintains this position on the mistaken belief that to do otherwise would be to concede its case against Vitalink. In any event, however, Bellcore's position is untenable. In DeGeorge, we noted that "[g]enerally, and in this case, the preamble does not limit the claims." Id. at 1322 n. 3, 226 USPQ at 761 n. 3. As the preceding discussion of our cases on claim preambles makes clear, this observation in DeGeorge can only have been de scriptive, rather than pre scriptive. We have long eschewed the use of an absolute rule according or denying all preambles limiting effect, having recognized that one cannot determine a preamble's effect except by reference to the specific claim of which it is a component. Claim 6, as drafted and in light of the specification, is plainly limited such that it literally reads only on methods that transmit packets having both source and destination addresses.

The district court further concluded that the above limitation regarding the contents of the packet compels one to construe the step of "associating with said packet an identifier" to require that the identifier be separate and distinct from the destination address; Vitalink urges us to affirm this construction of the claim. In other words, Vitalink would have us read the phrase "associating with said packet an identifier" as if it were "inserting into said packet a separate identifier." This construction of the claim is, however, no more tenable than Bellcore's. For, while nothing within Claim 6 considered in isolation impeaches the construction that Vitalink prefers, it is legal error to construe a claim by considering it in isolation. A claim must be read in view of the specification of which it is a part. Adams, 383 U.S. at 49, 86 S.Ct. at 713; Markman, 52 F.3d at 978-80, 34 USPQ2d at 1329-30. The specification of the '080 patent makes it clear to one of ordinary skill in the art that one can "associat[e]" an identifier with a packet within the meaning of Claim 6 either explicitly ( e.g., by the insertion of an additional "tree number" field into the packet) or implicitly ( e.g., by the insertion of a destination address field from which a tree number can be determined). Thus, the associating step of Claim 6 covers both an implicit or an explicit approach. The district court, by construing it to cover the explicit approach alone, read an additional limitation into the claim, an error of law. See, e.g., Specialty Composites v. Cabot Corp., 845 F.2d 981, 988, 6 USPQ2d 1601, 1606 (Fed. Cir. 1988).

Bellcore also contends that the district court erred when it interpreted the assigning step of Claim 6 to require that the source device assign the packet to be broadcast along one and only one spanning tree, on which the packet would travel to its destination without any mid-course changes in tree assignment. We, however, share the trial court's view of the assigning step's proper construction. The claim first recites the step of "assigning, by said source device, one of said trees to broadcast said packet." Although this clause appears to include no limitation regarding the possibility that midcourse changes in tree assignment might occur, the remainder of the claim renders such a construction of Claim 6 unworkable. First, the broadcasting step, according to which the packet is sent "through the system on said one of said trees," strongly suggests that the packet travels on the same tree, assigned at the outset, from its source to its destination. Second, and more importantly, the claimed method whereby the gateways execute their store-and-forward protocol precludes midcourse changes in tree assignment. According to the claim, each gateway receiving the packet "determin[es] for each said packet . . . said packet identifier" — that is, the identifier associated with the packet that is "indicative of said one of said trees" on which the packet has been broadcast. Each gateway then executes its first inhibit-or-forward decision according to whether it "process[es] packets having said identifier" — again, the identifier associated with the packet that is "indicative of said one of said trees" on which the packet has been broadcast. Thus, the identifier that indicates the "one of said trees" along which the packet is broadcast "through the system" remains the same throughout the packet's trip from its source to its destination. As a consequence, in the claimed method, the tree on which the packet travels must also remain the same: if the packet were assigned to a new tree mid-course, the gateways along that new tree would not forward it, inasmuch as they do not process packets having the identifier, indicative of a different tree, that was associated with the packet prior to its broadcast. The district court properly construed Claim 6 to require that the source device assign, prior to transmission, the one spanning tree along which the packet travels to its specified destination.

C. Summary Judgment of Noninfringement

The district court's decision to grant Vitalink's summary judgment motion flowed from its construction of the scope of Claim 6. Specifically, as we noted above, having construed the associating step of Claim 6 to require the insertion into the packet of a spanning tree identifier separate from the destination and source addresses, the court concluded that Bellcore could not make out an infringement case, literal or by equivalents, against Vitalink. Because that construction was erroneous, the court's reliance on the associating step as a basis for its summary judgment ruling was necessarily in error.

The district court also concluded that since the claimed method assigns the one spanning tree along which the packet travels to its specified destination and Vitalink's DLS system allows for mid-course corrections, Bellcore's literal infringement claim was legally defective as to the assigning step of Claim 6. Although the court's construction of the assigning step was correct, it does not necessarily follow that Bellcore's infringement claim against Vitalink must fail, because the record does not make it clear that Vitalink's DLS system never uses the claimed method. While we express no view on the ultimate question whether Vitalink's DLS system infringes Claim 6, either literally or by equivalents, it should be noted that any future infringement analysis respecting the assigning step should be undertaken with due attention to the principle that an accused product that sometimes, but not always, embodies a claimed method nonetheless infringes. See Paper Converting Mach Co. v. Magna-Graphics Corp., 745 F.2d 11, 20, 223 USPQ 591, 597 (Fed. Cir. 1984) ("[I]mperfect practice of an invention does not avoid infringement."); Roche Prods., Inc. v. Bolar Pharmaceutical Co., 733 F.2d 858, 861, 221 USPQ 937, 939 (Fed. Cir.) ("Section 271(a) [of Title 35] prohibits, on its face, any and all uses of a patented invention."), cert. denied, 469 U.S. 856, 105 S.Ct. 183, 83 L.Ed.2d 117 (1984).

The court's error in claim construction vitiates its conclusions as to Bellcore's ability to succeed on its infringement claims on the present limited summary judgment record. The declaratory judgment of noninfringement in favor of Vitalink must therefore be vacated.

CONCLUSION

Accordingly, we vacate the district court's summary judgment in Vitalink's favor and remand the case for proceedings on the issue of infringement, literal or by equivalents, consistent with this opinion.

VACATED AND REMANDED

COSTS

Each party to bear its own costs.