TARANTO, Circuit Judge.

Straight Path IP Group, Inc. owns U.S. Patent No. 6,108,704, entitled "Point-to-Point Internet Protocol," which describes certain protocols for establishing communication links through a network. On a petition for inter partes review filed by Sipnet EU S.R.O., the Patent Trial and Appeal Board cancelled claims 1-7 and 32-42 of the '704 patent based on determinations of anticipation and obviousness. Sipnet EU S.R.O. v. Straight Path IP Group, Inc., IPR 2013-246, 2014 WL 5144564 (PTAB Oct. 9, 2014). We now reject a claim construction on which the Board relied for its decision. We reverse the Board decision, and we remand for further proceedings under the correct construction.

BACKGROUND

The '704 patent identifies a deficiency in what the prior art taught about real-time voice or video communications between two processing units over a network, such as the Internet. According to the specification, the prior art disclosed successful protocols for such point-to-point communication between users and devices that maintained permanent network addresses. '704 patent, col. 1, lines 48-52. But for systems in which addressing is dynamic, i.e., in which devices obtain only temporary addresses on a network, "point-to-point communications in realtime of voice and video have been generally difficult to attain." Id., col. 1, lines 53-56. To solve the problem, the summary of the invention identifies a "point-to-point Internet protocol" that "exchanges Internet Protocol (IP) addresses between processing units to establish a point-to-point communication link," based on the first unit's querying "a connection server to determine the on-line status of" a second unit. Id., col. 1, lines 59-61, col. 2, lines 1-2. The summary also identifies a second protocol, which involves email signaling. Id., col. 2, lines 10-21.

The specification provides some details of operation—whose significance for claim construction is disputed, as discussed below. A processing unit, upon joining a network such as the Internet, automatically transmits its temporary network address and email address to a connection server. Id., col. 5, lines 25-29. The server stores the addresses in a database and timestamps them, id., col. 5, lines 29-31, thus "establish[ing]" the unit as "an active on-line party available for communication using the disclosed point-to-point Internet protocol," id., col. 5, lines 32-34; see id., col. 5, lines 35-38 (same for a second unit). To reduce the staleness of the status information, the server "may use the timestamps to update the status of each processing unit; for example, after 2 hours, so that the on-line status information stored in the database 34 is relatively current." Id., col. 5, lines 39-44. Another, seemingly even better means of keeping the database information accurate about true on-line status is this:

Id., col. 6, lines 6-16.

When a first unit seeks to set up a point-to-point communication link with a second unit, it "sends a query, including the E-mail address of the callee, to the connection server 26," which "searches the database 34 to determine whether the callee is logged-in by finding any stored information corresponding to the callee's E-mail address indicating that the callee is active and on-line." Id., col. 5, lines 55-60. "If the callee is active and on-line, the connection server 26 then performs the primary point-to-point Internet protocol; i.e., the IP address of the callee is retrieved from the database 34 and sent to the first processing unit 12." Id., col. 5, lines 60-64. The protocol does not include the actual establishing of the point-to-point communication, but once the IP address is sent to the first unit, the first unit "may then directly establish" communication with the callee using the latter's IP address. Id., col. 5, lines 64-67. And: "If the callee is not online when the connection server 26 determines the callee's status, the connection server 26 sends an OFF-LINE signal or message to the first processing unit 12." Id., col. 6, lines 1-4.

The specification then describes the "secondary point-to-point Internet protocol," which involves the sending of messages to an email server—either as a supplement to or independently of the "primary" protocol using the connection server. See, e.g., id. at col. 6, line 17, to col. 7, line 31. And it states that, using the described protocols, real-time point-to-point audio, video, and voice communication can "be established and supported without requiring permanent IP addresses to be assigned to either" the caller or callee. Id., col. 7, lines 32-36.

Claim 1 of the `704 patent is representative of the asserted claims:

`704 patent, col. 11, lines 2-22 (emphasis added to highlight the key claim phrase at issue).

In its petition for inter partes review of the `704 patent under 35 U.S.C. § 312, Sipnet requested cancellation of claims 1-7 and 32-42 as anticipated by and obvious over several prior-art references, most significantly "Net-BIOS" and "WINS." The Board, under authority delegated by the Director of the Patent and Trademark Office, instituted inter partes review under § 314. The Board then conducted the review pursuant to § 316 and reached a final decision cancelling the challenged claims under § 318. Sipnet, supra.

Of central importance on appeal, the Board adopted Sipnet's proposed construction of the claim language highlighted above. Although the parties agreed that the language requires "being on-line," they disagreed about whether, as Straight Path contended, the language refers to a present-tense status, J.A. 299-302, 305-10, or whether, as Sipnet contended, it "simply requires being registered with the server." Sipnet, 2014 WL 5144564, at *3. The Board adopted Sipnet's view as the broadest reasonable construction based on the specification: "`connected to the computer network' encompasses a processing unit that is `active and on-line at registration.'" Id. at *4. As is not disputed here, what the Board meant was that, to come within the query claim language, all the query from the first unit need do is request whatever the connection server has listed about a second unit's on-line status, even if the listed information is not accurate at the time of the query, i.e., even if it lists the second unit as online when, at that time, it is in fact not online.

Based on that construction, the Board concluded that claims 1-7, 32, and 38-42 were anticipated by NetBIOS, claims 1-7 and 32-42 were anticipated by WINS, and claims 33-37 were invalid for obviousness over NetBIOS and WINS. Id. at *8-15. Straight Path appeals under 35 U.S.C. §§ 141, 319, challenging the claim construction just described and also presenting an argument about the term "process" in the claims. We have jurisdiction under 28 U.S.C. § 1295(a)(4)(A).

DISCUSSION

A

There being no dispute here about findings or evidence of facts extrinsic to the patent, whether facts about outside-the-patent understandings of technical words or other facts, we conduct a de novo review of the Board's determination of the broadest reasonable interpretation of the claim language. See In re Cuozzo Speed Technologies, LLC, 793 F.3d 1268, 1279-80 (Fed.Cir.2015); Microsoft Corp. v. Proxyconn, Inc., 789 F.3d 1292, 1297 (Fed.Cir. 2015). Straight Path notes that the patent has now expired and on that basis asks us to determine the governing construction under the principles of Phillips v. AWH Corp., 415 F.3d 1303 (Fed.Cir.2005) (en banc), because the Board applies Phillips, rather than the broadest-reasonable-interpretation standard, for expired patents. See In re Rambus, Inc., 753 F.3d 1253, 1256 (Fed.Cir.2014). We need not explore the issues raised by that request, however, because we conclude that the Board adopted a claim construction that is erroneous even under the broadest-reasonable-interpretation standard.

We start with the claim language—which has a meaning that can only be called plain. The present tense "is" in "is connected to the computer network" plainly says that the query transmitted to the server seeks to determine whether the second unit is connected at that time, i.e., connected at the time that the query is sent. The question asked by the query is whether the device "is" connected, not whether it was connected or whether it is still registered as being connected even if that registration information is no longer accurate. It is not a reasonable interpretation of the claim language, considering its plain meaning, to say that it is satisfied by a query that asks only for registration information, regardless of its current accuracy.

The Board said nothing that either recognizes or disputes the plain present-tense meaning of the claim language on its face. Indeed, the Board's construction—"active and on-line at registration," Sipnet, 2014 WL 5144564, at *4 (emphasis added)—implicitly recognizes that being online is a status that can change over time: having the status "at registration" is having it at a particular time. The query required by the claim language asks if the callee "is" online, which is a question about the status at the time of the query. But the Board did not address the facially clear meaning, instead turning immediately to the specification.

Sipnet does much the same thing. Sipnet repeatedly recognizes and stresses the difference between "past online status" and "current online status," the latter being "opposed to the past status at registration." Sipnet Br. at 21 (emphases in original); see id. at 8-9. Yet Sipnet offers no argument that, as a matter of plain meaning, the claim language "is" calls for anything but present-status information. Nor does it point to anything in other claim language that contradicts that plain meaning. Like the Board, Sipnet relies entirely on the specification.

When claim language has as plain a meaning on an issue as the language does here, leaving no genuine uncertainties on interpretive questions relevant to the case, it is particularly difficult to conclude that the specification reasonably supports a different meaning. The specification plays a more limited role than in the common situation where claim terms are uncertain in meaning in relevant respects. The reason is that, unless there is a disclaimer or redefinition, whether explicit or implicit, the proper construction of any claim language must, among other things, "stay[] true to the claim language," and, in order to avoid giving invention-defining effect to specification language included for other descriptive and enablement purposes, "the court's focus remains on understanding how a person of ordinary skill in the art would understand the claim terms." Phillips, 415 F.3d at 1316, 1323, 1324 (internal quotation marks omitted); id. at 1321. Reflecting the distinct but related roles of the claims and specification, the governing approach to claim construction thus maintains claim language's key (not always decisive) role in claim construction: it stresses the importance of the specification in identifying and resolving genuine uncertainties about claim language, and in stating redefinitions or disavowals, id. at 1315-17, while it rejects a sequenced, dictionary-driven, burden-shifting approach to claim construction, id. at 1320-24. Under our Phillips approach, the plainness of the claim language necessarily affects what ultimate conclusions about claim construction can properly be drawn based on the specification. For that reason, the court has repeatedly stated since Phillips that redefinition or disavowal is required where claim language is plain, lacking a range of possible ordinary meanings in context. See Pacing Technologies, LLC v. Garmin Int'l, Inc., 778 F.3d 1021, 1024 (Fed.Cir. 2015) (citing authorities).

Here, the specification does not provide a basis for reasonably adopting a construction that contradicts the plain meaning of the claim language. The Board relied on just one passage from the specification—which says that "a second user operating the second processing unit, upon connection to the Internet through a connection service provider, is processed by the connection server to be established in the database as an active on-line party." `704 patent, col. 5, lines 34-38 (emphasis added); see id., col. 5, lines 31-34 (similarly as to the first unit: registration establishes it "as an active on-line party available for communication" using the disclosed protocol). But that passage says no more than that the unit is active and online—available for communication—at the time it registers. It does not expressly or implicitly redefine "is connected" to mean "is still registered, once was connected, and may or may not still be connected," and it does not otherwise establish that being active and online at the time of registration means, even if contrary to fact, being active and online when a caller's query for a callee's status comes in.1

Indeed, the immediately following passage recognizes the temporal nature of the status of actually being online. The specification says that the connection server "may use the timestamps to update the status of each processing unit" over time to try to keep the "on-line status information stored in the database 34 relatively current." Id., col. 5, lines 39-42. Whether that passage refers to "registration time-outs," Sipnet Br. at 23, or to "actively check[ing] whether a process is still connected to the network," J.A. 296 (Patent Owner's Response), the passage clearly presupposes that the database listing of a unit as an active online party can become false over time. Contrary to the Board's construction, the specification thus distinguishes, rather than equates, being online and being (or having been) registered.

The Board did not rely on any other basis for its construction, and Sipnet does not meaningfully do so in this court. Two additional passages from the specification are nevertheless worth noting. They confirm that there is no basis for departing from the plain meaning of the claim language.

The specification says that, when a first unit sends a query to the connection server, the latter "searches the database to determine whether the callee is logged-in by finding any stored information corresponding to the callee's E-mail address indicating that the callee is active and on-line." `704 patent, col. 5, lines 57-60. That language merely describes checking the database for stored information. It does not state that whatever information is stored, no matter how the connection server operates, establishes whether the callee is active and online.

Moreover, the specification immediately continues with a description of how a connection server might work so as to shrink if not completely eliminate any gap between recorded status and true status: "[w]hen a user logs off or goes off-line from the Internet, the connection server 26 updates the status of the user in the database; for example, by removing the user's information, or by flagging the user as being off-line." Id., col. 6, lines 6-9; see id., col. 6, lines 10-14 (the "server 26 may be instructed to update the user's information in the database by an off-line message . . . sent automatically from the processing unit of the user prior to being disconnected from the connection server"). At oral argument, Sipnet's counsel seemed to agree that the passage describes what must be "happening if the connection server answer is going to do what the claim language requires, supply an answer to the query whether the second process is connected to the computer network." Oral Argument at 24:10-25:03. The specification's indication of how a particular server process can provide accurate information undermines the notion that the specification generally redefines "is connected" to include active and online at registration, even if not at the time of the query.

The plain meaning of the claim language is therefore not overridden by the specification. And the plain meaning is positively confirmed by the prosecution history, which we have indicated is to be consulted even in determining a claim's broadest reasonable interpretation. See Proxyconn, 789 F.3d at 1298. In distinguishing claims 1-7 and 32-42 over NetBIOS and its "active name" disclosure to overcome a rejection during reexamination, the assignee of the `704 patent made the very distinction that is at issue here—between still being registered and actually being online:

Reply to Office Action of August 27, 2009, reexamination of `704 patent, control no. 90/010,416 (dated Nov. 27, 2009) at 11, 14-15. After the assignee made that distinction, the examiner withdrew the rejection and confirmed the claims.

One final point about this claim-construction issue: Sipnet suggests in various ways that the specification does not adequately describe or enable the systems or processes involving a query about current connection status under Straight Path's claim construction. But written-description and enablement challenges were not, and could not have been, part of the inter partes review that is now before us. See 35 U.S.C. § 311(b) (limiting challenges to prior-art challenges). Such challenges involve bottom-line or subsidiary factual issues that have not been litigated or adjudicated. Accordingly, Sipnet's arguments about insufficient specification support for the claims if they are given their plain meaning, arguments not adopted by the Board, do not alter our conclusion about claim construction. We offer no view on the merits of Sipnet's suggestion of written-description or enablement problems.

For the foregoing reasons, "is connected to the computer network" in the `704 patent's claims—and the counterpart claim phrases that the parties agree bear the same meaning—can only reasonably be understood to mean "is connected to the computer network at the time that the query is transmitted to the server." The Board did not apply this claim construction in considering the prior art, including NetBIOS and WINS. It should do so on remand.

B

Straight Path's second challenge to the Board's decision rests on the contention that the Board failed to construe "process." We hold that Straight Path did not preserve that contention. It did not request a construction of "process" in its preliminary response to Sipnet's petition to institute inter partes review, in its response after the Board instituted the review, or at the oral hearing before the Board. In particular, Straight Path never argued for a construction of "process" under which a process being connected meant something other than its host device being connected. Nor did Sipnet. The Board thus committed no error in not construing "process." Because Straight Path's "process"-based challenge depends entirely on its newly proposed construction, which it failed to preserve before the Board, this court does not address the challenge.

CONCLUSION

We reverse the Board's cancellation of claims 1-7 and 32-42 of the `704 patent, and we remand for further proceedings consistent with this opinion.

REVERSED AND REMANDED

DYK, Circuit Judge, concurring-in-part, dissenting-in-part.

I respectfully dissent from the majority's claim construction of the term "is connected to the computer network" to require absolute currency in a real-time assessment of connectivity. The majority's insistence that "is" requires absolute currency fails to take account of a common usage of the term "is" and our prior decision in Paragon Solutions, LLC v. Timex Corp., holding that a reference to "real-time" does not necessarily require absolute currency. 566 F.3d 1075, 1088 (Fed.Cir.2009). More importantly, the majority fails to give sufficient weight to the specification which Phillips v. AWH Corp. holds "is always highly relevant to the claim construction analysis" and is "the single best guide to the meaning of a disputed term." 415 F.3d 1303, 1315 (Fed.Cir.2005) (en banc) (internal quotations and citations omitted) (emphasis added).1

I

The patent here claims a system, an apparatus, and associated methods to facilitate real time communication between two users over the internet. When a user logs onto the internet, he is assigned a dynamic IP address (akin to a phone number for a computer). Because this IP address may be different every time the user logs on, direct communication is nearly impossible, just as making a phone call to someone would be if phone numbers changed on every call. The present invention purportedly solves this problem by maintaining a database of IP addresses in a central server. As the specification describes, when a user logs onto the network, his computer sends a current IP address and an email address to a central server. `704 patent, col. 5, lines 25-29. When another user wishes to initiate real-time communication with the first user (for example, a video telephone call), his computer will query the server with the first user's email address. Id. at col. 5, lines 55-60. If the database indicates that the first user is online, the server will send the second user the IP address of the first user, enabling the second user to initiate direct real-time communication with the first user. Id. at col. 5, lines 60-64. All parties agree that discovering whether a user "is connected to the network" involves a check of the server's database.

II

The present claim construction dispute arises in the context of an inter partes review in which the Patent Trial and Appeal Board ("the Board") found the claims of the patent anticipated and/or obvious over two pieces of prior art that disclosed similar server-based databases of IP addresses. The patentee contends that these prior art disclosures do not read on the patent claims because the claims here require a real-time check, which is not disclosed by the prior art. Relying on the specification, which indicates that the database only need be "relatively current," `704 patent, col. 5, lines 39-42, the Board rejected this construction. On appeal, the entire anticipation/obviousness question now depends on the construction of the term "is connected to the network."

The majority rejects the Board's specification-based construction, holding that the word "is" necessarily denotes currency and cannot accommodate a situation in which information is even slightly out of date. The majority states that "the present tense `is' in `is connected to the computer network' plainly says that the query transmitted to the server seeks to determine whether the second unit is connected at that time, i.e., connected at the time that the query is sent." Maj. Op. at 1360. "When claim language has as plain a meaning on an issue as the language [in this claim] does, . . . [t]he specification plays a more limited role than in the common situation where claim terms are uncertain in meaning in relevant respects." Id. at 1361. The majority then concludes that "the specification does not provide a basis for reasonably adopting a construction that contradicts the plain meaning of the claim language." Id. at 1361. According to the majority, the Board's construction was erroneous because it "did not address the facially clear meaning, instead turning immediately to the specification." Id. at 1360. In other words, the majority relies on the meaning it assigns to claim language based on its own knowledge of word usage rather than relying on the patentee's own specification.

III

Contrary to the majority's assertion, ordinary usage easily accommodates the Board's interpretation of "is connected." If a person says that "John is at home," this might lead to the question: "How do you know?" The response "I spoke to him five minutes ago" would not be viewed as contradicting the original statement, even though John might have left home in the intervening five minutes. In other words, the use of the word "is" does not necessarily imply absolute accuracy or absolute currency.

In any event, under the Phillips approach, we must look to the specification as the "single best guide to the meaning of a disputed term." Phillips, 415 F.3d at 1315. This is true, contrary to the majority's assertion, even in cases where language, on its face, appears to have a plain meaning, because, as Phillips states, the specification "is always highly relevant to the claim construction analysis." Id. (emphasis added). The only meaning that matters is the meaning in the context of the patent. See id. at 1316 (citing to and quoting Netword, LLC v. Centraal Corp., 242 F.3d 1347, 1352 (Fed.Cir.2001) ("The claims are directed to the invention that is described in the specification; they do not have meaning removed from the context from which they arose."))

In other cases, we have appropriately paid primary attention to the specification, including a case involving a nearly identical situation, Paragon Solutions, 566 F.3d at 1088. The majority's insistence that "is" requires currency independent of the specification is directly contrary to this prior decision, which faithfully followed Phillips. We held that "displaying real-time data" "cannot mean instantaneous" based on a reading of the claim language in the context of the specification. Paragon Solutions, 566 F.3d at 1088. This case involved an exercise monitoring system (which can be worn on the wrist) that displays data to the user from both an "electronic positioning device" and from a "physiological monitor." Id. at 1083. Despite seemingly similar clarity in the language of the claim—"a display unit configured for displaying real-time data"—we recognized that claim language cannot be read in isolation and looked to the specification in determining the ordinary meaning of the claim language. Id. at 1088. Because the claim required data that, according to the specification, took time to develop, we rejected the district court's technical dictionary-based construction and held that "real-time" did not mean "instantaneous." See id. at 1092. The majority's conclusion that "is" necessarily requires currency here contravenes the holding of Paragon Solutions.

The majority's construction of the term is also contrary to the specification in this case. The specification is clear that "is connected" does not require a real-time check. The specification—the "single best guide to the meaning of a disputed term," see Phillips, 415 F.3d at 1315—describes that, when a query is received, the server "searches the database . . . to determine whether the callee is logged-in by finding any stored information corresponding" to that queried user. `704 patent, col. 5, lines 57-59 (emphasis added). This information in the database, as described by the specification, is kept "relatively current." Id. at col. 5, lines 39-42. Checking historical "relatively current" information in a database is not a "real time" determination.

The majority's construction is not only inconsistent with the general description of the invention, it is also inconsistent with the described embodiments. At oral argument, the patentee conceded that its construction, now adopted by the majority, would require the database to be always accurate.2 In the preferred embodiment, a database record is created when a user logs on to the network. `704 patent, col. 5, lines 25-29. Everyone agrees that checking whether a user "is connected to the network" involves checking that database record. But this embodiment includes mechanisms to detect if a user is no longer on-line that do not guarantee the accuracy of that record. For example, the server will periodically (e.g., every 2 hours) update the database, looking to see whether a record has been updated during that time period. Id. at lines 39-40. If the record is too old, the server will change the database record to indicate that the user is no longer online. See id. This check can be out of date because of the time between periodic checks. The embodiment also describes that when a user logs off, that user's computer can send a log-off message to the server, which will then update the database record to indicate that that particular user is not online. Id. at col. 6, lines 6-16. As the Board found, even this method will not prevent a non-current response to a query to the database when, for example, the first user's computer crashes or otherwise fails to send a log-off message to the server. Thus, as the patent itself recognizes, the information in the database will only be "relatively current," and there is no disclosure in the specification that would warrant construing "is connected" to require absolute accuracy.

I respectfully dissent.