US 7643481

I'll verify the patent identifiers against live sources before summarizing.

Now let me retrieve the actual claim text and check for any litigation.

Verification status first

I ran targeted searches against Google Patents, FreePatentsOnline, Justia, Patents-Review, and general docket/web sources. I was able to confirm the bibliographic record and the abstract, but I was not able to retrieve the granted claim text for US 7,643,481 from the sources returned. Rather than reconstruct claim language, I flag that gap explicitly below. The full patent text supplied to me in this session is also truncated before the claims.

---

US 7,643,481 B2 — bibliographic record

Family / priority notes (as reported by the sources, taken literally):

• Google Patents shows priority claimed from US 09/343,409 (→ US 6,335,932 B2) and US 09/527,999 (→ US 6,996,099 B1).
• The patent-application copy states 11/065,562 is a continuation of 09/527,999 (filed 2000-03-17) and a continuation-in-part of 09/343,409 (filed 1999-06-30) and 09/527,999, with 09/527,999 claiming priority to provisional applications 60/124,878 (1999-03-17), 60/135,603 (1999-05-24), and 60/149,706 (1999-08-20).
• Note: US 6,996,099 carries the same title, "Network switch having a programmable counter," and the same inventors — it is the parent in this chain, not a separate invention.

---

Abstract (verbatim)

"A network switch for switching packets from a source to a destination includes a source port for receiving an incoming packet from a source, a destination port that contains a path to a destination for the packet, and a programmable counter unit for counting a number of packets of selected packet types which are received by the switch."

The specification's own summary paragraph repeats this language almost verbatim, which is the strongest signal available to me about the intended scope of the head claim.

---

Independent claims — I cannot give you verified claim text

I could not obtain the granted claims of US 7,643,481 from the sources returned in my searches. I therefore will not paraphrase or reconstruct independent claims, because doing so would risk fabricating claim scope — the single most consequential error possible in this kind of analysis.

What is genuinely grounded from the materials I do have:

• The abstract and the specification summary both frame the invention as a network switch having (a) a source port receiving an incoming packet, (b) a destination port containing a path to the destination, and (c) a programmable counter unit for counting a number of packets of selected packet types received by the switch. Given that claim drafting in the issued patent family mirrors this language, I would expect at least one independent claim to track it — but I state that as an expectation, not a finding.
• The surrounding specification describes the larger Broadcom switch-on-chip context, including MIB counters, programmable registers, status/control/config registers, rules tables, and a fast filtering processor (FFP) that classifies packets by field (L2–L7) and applies actions such as discard, CPU redirect, mirrored port, priority mapping, and TOS modification. These are the architectural elements from which a programmable packet-type counter would plausibly be programmed, but the specification's presence of these elements does not establish that they are recited in any claim.
• The examiner cited prior art includes US 5,987,507 (Creedon et al.), "Multi-port communication network device including common buffer memory with threshold control of port packet counters," and US 6,263,368 (Martin), "Network load balancing for multi-computer server by counting message packets to/from multi-computer server" — both counter-related, which is consistent with a counter-focused claim set.

Recommendation for a definitive answer: pull the claim set directly from USPTO PatentCenter (application 11/065,562), the Google Patents "Claims" tab for US7643481B2, or the granted patent PDF (US007643481). Those are the authoritative sources; I have not been able to reach them in this session.

---

Litigation check (USPTO / CAFC, "2026 dockets")

I searched for CAFC and district-court activity naming US 7,643,481 and found no case, docket entry, or opinion involving this patent number. The 2026 Federal Circuit activity involving Broadcom that did surface is about entirely different patents:

• TexasLDPC Inc. v. Broadcom Inc., No. 25-1074 (Fed. Cir. Sept. 14, 2026) — LDPC error-correction patents (US 8,418,023; 8,555,140; 9,112,530; 8,359,522; 8,656,250), not the '481 patent.
• Caltech v. Broadcom appellate briefing — error-correction patents (e.g., 7,916,781), not the '481 patent.

This is a negative search result, not proof of nonexistence. This patent is a 1999-priority Broadcom switch patent that went abandoned-for-failure-to-pay maintenance fees effective 2023-02-22, which makes it an unlikely infringement-assertion candidate. I have not confirmed the full reassignment chain (Broadcom → Avago Technologies General IP (Singapore) → Avago Technologies International Sales Pte. Ltd.) beyond the Google Patents reassignment entries, and I have not independently verified PTAB activity for this number.

---

Discrepancies I am flagging rather than silently correcting

• One aggregator (idiyas) renders the number as "US 7643481 B1." That is a kind label error; the grant publication is B2. I am reporting the discrepancy rather than treating the aggregator's label as controlling.
• Google Patents' "Info" table lists an unrelated reference, US6556205A, in a cell adjacent to the application number. That appears to be a rendering artifact of the page, not a data element of this patent.
• FreePatentsOnline lists the assignee as Broadcom Corporation (Irvine, CA) and the pre-grant publication date as 2010-01-05; the correct reading is that 2010-01-05 is the grant date and the pre-grant publication was 2005-06-30.

Bottom line: title, assignee, inventors, filing date, issue date, abstract, and status are confirmed with high confidence. The independent-claim overview is not something I can supply responsibly from what I was able to retrieve, and I am declining to guess at it.

I'll search for litigation involving US patent 7643481 across patent litigation databases.

Let me run more targeted searches on patent litigation databases.

Let me try more specific searches on litigation databases and Broadcom's known cases.

Litigation Report — U.S. Patent No. 7,643,481

"Network switch having a programmable counter"
Inventors: Shiri Kadambi; Shekhar Ambe
Original assignee: Broadcom Corporation (Irvine, CA)
Application No. 11/065,562; Continuation of Ser. No. 09/527,999 (now U.S. 6,996,099)
Priority: 1999-03-17 | Filed: 2005-02-25 | Granted: 2010-01-05
Current assignee of record: Avago Technologies International Sales Pte Ltd (via Broadcom → Avago → Broadcom merger chain)
Legal status (Google Patents snapshot): "Expired – Fee Related," adjusted expiration 2023-02-22

---

Bottom line

I did not find any litigation — district court, Federal Circuit, ITC, or PTAB — in which U.S. Patent No. 7,643,481 is identified as a patent-in-suit or challenged patent.

I cannot responsibly tell you that no such case exists. My searches returned no case records naming this patent, but my access was limited to general web/patent-database search rather than the closed litigation dockets (PACER/CM-ECF) or commercial analytics platforms that would give a definitive answer. See "Limits of this search" below.

---

Case table

No IPR, PGR, CBM, or reexamination proceeding directed to the '481 patent was identified either.

---

What I checked, and the negative results

1. Search on "7643481" and "7,643,481" plus litigation terms — returned no case documents naming the patent.
2. Broadcom's own enforcement campaigns (checking whether '481 appeared as an asserted patent):

• Broadcom v. Emulex (C.D. Cal., filed Sept. 2009) — asserted patents per Broadcom's own 10-K disclosure were U.S. 7,058,150; 7,471,691; 7,450,500; 6,424,194; 7,486,124; 7,724,057. The '481 is not among them.
• Broadcom v. Qualcomm (C.D. Cal./D.N.J., 2005 onward) — I found no record of '481 being asserted.
• Broadcom v. Netflix (N.D. Cal. 3:20-cv-04677) — asserted patents included 7,121,387; 6,982,663; 8,572,138; and the '121 patent. The '481 is not among them.
• Broadcom v. CSIRO (E.D. Tex., 2009) — single-patent declaratory judgment action; not '481.
• In re Certain Infotainment Systems (ITC Inv. 337-TA-1119, Broadcom complainant) — asserted U.S. 7,437,583; 7,512,752; 7,530,027; 8,284,844; 6,937,187; 8,902,104. The '481 is not among them.

1. Post-acquisition assertion by PAEs. I found no record of the '481 being asserted by a third party after the Avago/Broadcom acquisitions. (By contrast, the sibling/family patents in Broadcom's switch portfolio have historically been asserted by Broadcom itself, not by NPEs, and this patent expired 2023-02-22, which sharply narrows any window for late-life assertion.)

---

False positives I had to discard (read carefully)

The number "7643481" is easily confused with other strings. The following are not this patent and must not be reported as litigation over it:

• U.S. 9,095,348 ("the '348 patent") — Hologic, Inc. v. Minerva Surgical, Inc., D. Del. No. 1:15-cv-01031-SLR (and IPR2016-00680/685). Different patent, different owner, different technology (endometrial ablation).
• A "7643481" Lithuanian and Brazilian registry entries (a company notice number in the Lithuanian registry; a Brazilian commercial-registration row). Not patents.
• Australian application AU 76434/81 (granted as AU 542655B2, "Combined book leaf holder and bookmark") — the string "AU7643481A" appears in its INPADOC family, an artifact of assembling "AU" + "76434/81."

One further note on a data-quality discrepancy I observed but will not "correct" silently: the third-party site idiyas.com renders this patent as "US 7643481 B1." The authoritative Google Patents and FreePatentsOnline records both show US 7,643,481 B2. The B2 kind code is correct.

---

Why the absence is plausible

• The '481 is a continuation of U.S. 6,996,099, a member of Broadcom's late-1990s/2000-era switch-on-chip family (priority 1999-03-17). Broadcom historically enforced other members of this family (e.g., the '150, '691, '500 patents against Emulex; the '583, '752, '027 patents at the ITC) rather than this particular continuation.
• The patent expired 2023-02-22, so no infringement action could be filed after that date, and the enforceability window was 2010–2023.
• The '481 does not appear in Broadcom's disclosed "Intellectual Property Proceedings" narrative in its periodic SEC filings that I reviewed.

Limits of this search — please verify independently

My searches did not include the authoritative, paywalled/closed dockets. Before relying on a negative finding, I recommend confirming against:

• PACER / CM-ECF and CourtListener RECAP (docket-level search by patent number)
• Docket Navigator and Lex Machina (patent-centric litigation analytics)
• Unified Patents Portal — https://portal.unifiedpatents.com/litigation/caselist (and its per-patent "Litigation" tab at .../patents/patent/US-7643481-B2)
• RPX Insight (insight.rpxcorp.com) — per-patent litigation listing
• USPTO Patent Center / Global Dossier "Litigation" and "Assignment" tabs for the patent, and MaxVal / Reed Tech patent litigation feeds
• PTAB Endpoint / PTAB API for any IPR/PGR/CBM against the '481
• Federal Circuit docket search for appeals naming the '481

If you can give me access to or the results of a Docket Navigator or Lex Machina query for 7,643,481 specifically, I can draft the case table with plaintiff, defendant, jurisdiction, case number, filing date, and disposition for each case.

---

A caution on the premise

You asked me to prioritize search results over training data where they conflict. Here the search results did not contain any litigation record for this patent — there was no conflict to resolve. I am therefore reporting an absence of evidence rather than evidence of absence, and I have flagged exactly which sources were and were not consulted so the conclusion can be verified.

I'll start by checking whether any AIA trial proceedings exist for this patent, since the structured block indicates none are indexed.

Let me run more targeted searches to be certain no AIA proceeding exists on this specific patent.

Let me do final confirmation searches, including the related family members and any Broadcom assertion campaigns.

I ran the canonical structured source (USPTO ODP via the "PTAB proceedings on file" block) plus four independent web searches (patent number + IPR/PGR/CBM; patent number + "Final Written Decision"/"institution"; Broadcom + 7643481 + validity challenge; and the Google Patents / FreePatentsOnline records). No AIA trial proceeding involving US 7,643,481 surfaced in any of them.

---

Proceedings overview

Total AIA trial proceedings on file for US 7,643,481: 0 — 0 active, 0 claims invalidated, 0 claims sustained, 0 settled, 0 institution denials. Because the PTAB has never construed a single claim of this patent, a defendant today faces an untested patent, not a hardened one: there is no FWD to point to, no cancellation certificate narrowing the claim set, and no § 315(e)(2) estoppel to exploit — but equally, no adverse PTAB precedent constraining the patent owner's claim-construction or validity positions either.

No proceedings to enumerate

The structured ODP block returns no AIA trials, and my independent searches corroborate that:

• No IPR, PGR, or CBM petition bearing the number 7,643,481 appears in the PTAB petition/decision corpora that surfaced in search.
• The Google Patents legal-events record for US7643481B2 (source: https://patents.google.com/patent/US7643481B2/en) contains only prosecution, assignment, and security-interest events — no ipr, pgr, cbm, or appeal legal-event entries. For a patent that has been through an AIA trial, Google Patents normally carries the trial's legal events.
• The FreePatentsOnline record (https://FreePatentsOnline.com/7643481.html) shows only the ordinary patent family and cited references, with no third-party challenge noted.

Per your operating rules, I am not inventing proceeding numbers to fill the template. There is nothing to report at the per-proceeding level, and I will not manufacture a "7643481 — Petitioner v. Broadcom" heading.

Confidence and caveat. I am highly confident there is no AIA trial that reached a Final Written Decision on this patent, since FWDs are published and indexed. I am somewhat less confident about a petition that was denied institution in the 2012–2016 window, since pre-2018 non-institution decisions are less consistently indexed by third-party sources — though the ODP block is the authoritative feed and it is empty. The definitive check is a patent-number search in PTAB E2E: https://ptacts.uspto.gov/ptacts/ (search "7643481" in the Application/Patent Number field, all trial types, all years). If that search is empty, the count is zero.

---

Adjacent activity worth knowing (NOT proceedings on this patent)

These are not proceedings on US 7,643,481 and should not be cited as such. They are relevant only as pattern evidence about the parties.

• Netflix, Inc. v. Avago Technologies International Sales Pte. Ltd., IPR2021-00542 — challenges to claims 11, 12, 14, 22, and 24 of US 8,572,138 (a different Broadcom/Avago patent), arising out of Broadcom Corp. et al. v. Netflix, Inc., No. 3:20-cv-04677-JD (N.D. Cal.). Broadcom and Avago asserted a twelve-patent portfolio against Netflix; the petition I reviewed addresses the '138 patent only. I could not verify whether US 7,643,481 was among the patents asserted in that campaign, and I found no petition challenging it.
• Wi-Fi One, LLC v. Broadcom Corp. — a line of IPRs (including on US 7,457,215) in which Broadcom was the petitioner, ultimately producing the en banc § 315(b) time-bar decision, Wi-Fi One, LLC v. Broadcom Corp., 878 F.3d 1364 (Fed. Cir. 2018), and the Supreme Court's Thryv, Inc. v. Click-To-Call Technologies, LP, 140 S. Ct. 1362 (2020). This shows Broadcom's offensive IPR posture, not a defensive problem for this patent.
• Family members of the '481 include US 6,996,099 and US 6,335,932 (the '099 shares the exact title "Network switch having a programmable counter"; the '932 is the external-priority parent of this continuation chain, per the Google Patents priority record). I found no AIA trial on either, but I did not verify those separately to the same depth.

No Unified Patents or other defensive-aggregator filing on the '481 surfaced.

---

Strategic summary

Claim status. All claims of US 7,643,481 are UNTESTED at the PTAB. None are canceled; none are sustained; none have been construed by the Board. This is categorically different from the "hardened patent" and "dead claim" poles you asked about — there is simply no administrative record. Practically, that means every claim the patent owner chooses to assert in a demand letter is still live, and you have no PTAB cancellation to wave at the sender. The corollary is that the patent owner has no PTAB win to wave at you either.

Estoppel landscape. Because no IPR was ever instituted, no § 315(e)(2) estoppel exists against anyone. There is no petitioner, no privy, no RPI, and no instituted ground. For a defendant being asserted against today, every prior-art ground remains fully available — § 102, § 103, and § 112 — in district court, the ITC, or an ex parte reexamination. Nothing has been "used up." Conversely, this also means no petitioner has had grounds denied institution in a way that would forfeit them for you; you have a clean slate.

A timing issue dominates everything else. The Google Patents record states the term carried an adjusted expiration of 2023-02-22, with legal status "Expired – Fee Related." If that is right, US 7,643,481 is no longer enforceable for prospective conduct. That changes the defensive playbook substantially: (a) any infringement theory is limited to past damages within the six-year lookback of 35 U.S.C. § 286 running backward from the complaint date; (b) an IPR petition is legally still possible against an expired patent, but your practical leverage is in the damages case, not in obtaining prospective relief; and (c) you should check whether the § 286 window has already closed on some or all of the accused period. Verify the expiration independently against the USPTO Patent Center "Patent Term Adjustment/Extension" record before relying on it — an expiration calc that turns on a 1999–2001 priority chain and terminal disclaimers is exactly the kind of thing worth confirming from the source of truth rather than a third-party abstract.

Pattern signals. None on this patent: no repeat petitioner, no PO appeal history, no aggregator. The absence of any IPR over a 27-year life span is itself informative. Broadcom-family switch patents were heavily asserted in the 2015–2021 era (Broadcom/Avago v. Netflix; Broadcom's own IPR campaign against Wi-Fi One), and a widely asserted, commercially significant patent in that portfolio would normally have drawn at least one petition. The likeliest explanations are that the '481 was never asserted on a stand-alone basis, that it was asserted only in portfolios subject to quick settlement or licensing, or that it was perceived as too narrow or too close to expiration to be worth a petition. Do not over-read it: an unasserted, un-IPR'd patent is not evidence of invalidity — it is evidence of no information.

---

Recommended next steps

• Don't tell your client the claims are dead. They are not. Say: "No PTAB proceeding has ever been instituted on this patent; there is no cancellation certificate and no estoppel on file."
• Confirm the null result at the source before you rely on it. Search https://ptacts.uspto.gov/ptacts/ for "7643481" across IPR, PGR, and CBM. For completeness, also check the PTAB "Patent Trial Proceedings" tab on the Google Patents page and the USPTO Patent Center transaction history for any ex parte reexamination — I did not find one, but I did not exhaustively verify it.
• Verify the expiration date. Pull the Patent Center term record for US 7,643,481. If 2023-02-22 is correct, redirect the defense to (i) § 286 past-damages exposure and (ii) marking/§ 287 notice and laches-type arguments, and confirm whether any accused conduct post-dates expiration.
• If the patent is still enforced against you, you retain a full IPR menu. No § 315(b) one-year clock has ever started for a proceeding, so the only bar date is the one triggered by your own service of a complaint (35 U.S.C. § 315(b)). If you were served more than one year ago and are a real party in interest, you are time-barred from filing; if you want PTAB review, coordinate with an unserved/independent party or file ex parte reexam (which has no § 315(b) bar and can be filed anytime, including against an expired patent).
• Nothing to link as an FWD. There is no Final Written Decision, no institution decision, and no Federal Circuit appeal for this patent. If you need the authoritative negative, cite the empty ODP AIA-trial record and the PTAB E2E patent-number search result — that is the best available evidence of the absence, and it is a stronger cite than any secondary source.

Bottom line for a defendant: you are facing an unlitigated-at-the-PTAB, apparently expired 1999-priority network-switch patent. There is no claim-level PTAB outcome to rely on in either direction. Your defense rests on the expiration/damages math and on original prior-art work you will have to do yourself — not on borrowing someone else's IPR result, because there isn't one.

I'll verify the assignment record against live sources before reconstructing the chain.

Ownership & Assignment Analysis — US 7,643,481 B2

Scope note. Bibliographics were established in the prior section and are not restated. This section reconstructs the assignment chain. The only reel/frame number I can state with grounding is 047195/0827, which appears in the Google Patents legal-events text itself (discussed below). The USPTO Assignment Center result pages at https://assignmentcenter.uspto.gov/ and https://assignment.uspto.gov/patent/index.html were not reachable with per-patent reel/frame detail in this session, so every other reel/frame is marked not retrieved rather than estimated. I have not reconstructed any assignment that I could not see in a source.

---

Inventors

Pattern notes.

• Only two inventors are named on a disclosure of this breadth (an entire switch-on-chip architecture). That is normal for the Kadambi switch-SOC family: the same two names recur across the sibling patents already surfaced — US 6,335,932, US 6,850,521, US 6,707,817, US 7,184,441, US 7,720,055, and the same-titled parent US 6,996,099 ("Network switch having a programmable counter," Kadambi et al., Feb. 2006). This is a coherent engineering-team family, not a hired-inventor or acquire-and-refile pattern.
• Departure timing within 12 months of filing: not determinable. I found no employment-change, non-compete, or departure record for either inventor tied to this docket. I am explicitly not inferring a fire-sale precursor from inventor movement, because I have no evidence of inventor movement.
• No inventor-to-Broadcom assignment record for the 2005 CIP surfaced in the legal-events data retrieved. Such records almost certainly exist for the 1999 parents, but I could not retrieve their reel/frame. Gap flagged, not filled.

---

Original assignee

Broadcom Corporation, Irvine, CA — a California corporation, named on the face of the granted patent and confirmed by both FreePatentsOnline and the Google Patents record.

• Line of business: fabless semiconductor design; Ethernet/network switching silicon and SoC products were a core Broadcom business at the time of filing. The patent's own specification describes a switch-on-chip (SOC 10) with EPIC/GPIC/IPIC modules, a CPS channel, MMU, CBP/GBP memory hierarchy and a fast filtering processor — i.e., the architecture of a commercial switching product line, and Broadcom did commercialize Ethernet switch silicon.
• Did they ship a product embodying the claims? Broadcom shipped switch-on-chip silicon consistent with this disclosure, but I cannot verify from the record that any specific shipped part practices the granted claims, because the granted claim text could not be retrieved in this session (see the prior section's flag). I am not asserting product-to-claim correspondence.
• Current status: Broadcom Corporation ceased to be an independent public company on 2016-02-01, when Avago Technologies Limited's acquisition completed (mirrored in the Avago successor's Form 10-Q: "We completed the Broadcom Transaction on February 1, 2016"; the surviving parent renamed itself Broadcom Limited and later Broadcom Inc.). Broadcom Corporation continues to exist as a California corporation and an indirect subsidiary of the Broadcom group. It is not in bankruptcy, not dissolved.
• Assertion posture of the original assignee: Broadcom has asserted switch patents against competitors historically — e.g., Broadcom Corp. v. Qualcomm Inc., No. 8:05-cv-00467 (C.D. Cal.), where the court addressed permanent-injunction relief in the chip market. That litigation activity is documented, but I found no evidence that US 7,643,481 was asserted in that or any other case, so it is context, not a finding about this patent.

---

Assignment timeline

All entries below are recordation events as reported in the Google Patents legal-events record for US 7,643,481; execution dates are not exposed by that source and are marked as such. Where I say "correspondent: not retrieved," the Assignment Center cover-sheet data (attorney of record for the recording) was not reachable — I will not guess at a firm name.

1. 2016-02-11 (recorded; execution date not retrieved) — Reel/frame: not retrieved

• Conveyance: Patent Security Agreement (grant of security interest — not an ownership transfer)
• Assignor: Broadcom Corporation
• Assignee: Bank of America, N.A., as Collateral Agent
• Correspondent: not retrieved
• Context: securitization — collateral grant under Broadcom's 2016 credit facility (the "2016 Credit Agreement" tied to the February 1, 2016 Broadcom/Avago combination, described in the successor's Form 10-Q as consisting of 2016 Term Loans and a 2016 Revolving Credit Facility, with Bank of America as administrative agent). Recorded as a collateral agent entry, so it is a lien, not a change of owner.

2. 2017-02-01 (recorded; execution date not retrieved) — Reel/frame: not retrieved

• Conveyance: Assignment of Assignor's Interest (full ownership transfer)
• Assignor: Broadcom Corporation
• Assignee: Avago Technologies General IP (Singapore) Pte. Ltd.
• Correspondent: not retrieved
• Context: internal corporate reorganization — intra-group consolidation of Broadcom Corporation's U.S. patent estate into the group's Singapore IP-holding entity following the February 2016 combination.

3. 2017-02-03 (recorded; execution date not retrieved) — Reel/frame: not retrieved

• Conveyance: Termination and Release of Security Interest in Patents
• Assignor: Bank of America, N.A., as Collateral Agent
• Assignee: Broadcom Corporation
• Correspondent: not retrieved
• Context: release of lien / internal housekeeping — extinguishes the entry-1 security interest. "Assignee" here is the party receiving back the encumbered title; it is a release, not a purchase.

4. 2018-10-04 (recorded; execution date not retrieved) — Reel 047195 / Frame 0827 (see reading caveat)

• Conveyance: Merger
• Assignor: Avago Technologies General IP (Singapore) Pte. Ltd.
• Assignee: Avago Technologies International Sales Pte. Limited
• Correspondent: not retrieved
• Context: internal corporate reorganization — the IP-holding entity was merged into the group's international sales entity; ownership never left the Broadcom group.
• Reading caveat, stated explicitly: the string "REEL: 047195 FRAME: 0827" appears in this patent's record only as the target of the next entry's corrective assignment ("…PREVIOUSLY RECORDED AT REEL: 047195 FRAME: 0827"). I read 047195/0827 as the reel/frame of the merger record itself, but I could not open the Assignment Center record page to confirm the frame's contents. Treat it as high-probability, not verified.

5. 2018-11-05 (recorded; execution date not retrieved) — Reel/frame: not retrieved

• Conveyance: Corrective Assignment (to correct the effective date of the merger previously recorded at reel 047195/frame 0827)
• Assignor: Avago Technologies General IP (Singapore) Pte. Ltd.
• Assignee: Avago Technologies International Sales Pte. Limited
• Correspondent: not retrieved
• Context: administrative correction of entry 4's effective date — the same single correspondent would have filed both; I could not retrieve the correspondent name, so I cannot test that.

6. 2023-02-22 — no reel/frame (not an assignment)

• Conveyance: none — adjusted expiration due to non-payment of maintenance fees (Expired – Fee Related)
• Context: patent lapsed; no transfer event.

Chain ownership end-state: Avago Technologies International Sales Pte. Limited (a Broadcom Inc. group entity), holder of an expired patent.

Not in this chain (worth stating affirmatively): no assignment to any entity whose name or structure matches an NPE vehicle; no inventor-assignment record surfaced for the 2005 CIP; no recorded license, change-of-name, or second security agreement on this patent number in the legal-events record.

---

Timeline diagram

timeline
    title Ownership of US 7643481
    1999 : Priority applications filed with Broadcom
    2005 : CIP application filed by Broadcom Corporation
    2010 : Patent granted to Broadcom Corporation
    2016 : Security interest recorded to Bank of America
    2017 : Security interest released back to Broadcom
         : Assigned to Avago Technologies General IP
    2018 : Merged into Avago Technologies International Sales
         : Corrective assignment recorded
    2023 : Patent expires for unpaid maintenance fees

---

NPE / troll-pattern signals

1. Shell-entity transfer — not present. Every assignee in the chain is a named corporate entity inside one corporate group: Broadcom Corporation (operating, Irvine CA) → Avago Technologies General IP (Singapore) Pte. Ltd. → Avago Technologies International Sales Pte. Limited. There is no "IP / Patents / Licensing / Holdings / Ventures"–suffixed LLC in the chain, no registered-agent service address, and no single-member Delaware or Texas LLC. Caveat I am stating rather than hiding: "General IP (Singapore) Pte. Ltd." is an IP-holding subsidiary — that is a structuring vehicle. But the tell that separates it from an NPE is present in the record: it was the assignee of a same-group transfer (entry 2, 2017-02-01) with the encumbering bank lien released two days later (entry 3, 2017-02-03) — housekeeping inside a group, not a sale to an assertion vehicle.

2. Known asserter in the chain — not present. No assignee in this record matches any entity on the requested lists: Acacia Research, Marathon Patent Group, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio IP Ventures, MPHJ, Lumen View, Round Rock Research, Document Generation Corp, or any Erich Spangenberg entity. No RPX or Unified Patents asserter-directory entry naming US 7,643,481 as an asserted patent surfaced in the sources returned this session.

   • Data-quality flag: the Unified Patents portal page that lists US-7643481-B2 renders its assignee field as "Zhong Sheng." Zhong Sheng is an inventor on an unrelated cluster of switch patents shown on the same page; this is a field-parse artifact, not an assignee. The same page also gives a priority date of 1999-03-16, which conflicts with the 1999-03-17 priority date in the patent's own family data. Following the authoritative full text supplied to me, 1999-03-17 governs.

3. Repeat correspondent across the chain — unclear / not assessable. This is the single most informative field for the task and it is the one I could not obtain. No correspondent (recording attorney, firm, or address) was retrievable for any of the five recorded events. Because the assignment coversheets could not be opened, I cannot test recurrence, and I decline to name a firm. If you need this: open each reel/frame in Assignment Center and read the "Correspondent" block — for the 2018 pair, entry 4 (047195/0827) and entry 5 (corrective) will almost certainly show the same correspondent, which is expected housekeeping rather than an NPE tell.

4. Cascading transfers — not present. Two ownership events across a 20-month window (2017-02-01; 2018-10-04), nineteen years after filing, both intra-group, with a lien release interpolated between them (2017-02-03). There are no chained LLC assignees, no shared correspondent address to compare (field unavailable), and no common-principal shell structure. Note the one adjacent data point that would matter if it recurred: the 2018 pair (merger + corrective assignment of the same transaction) is a single transfer recorded twice, not two transfers — do not count it as cascading.

5. Pre-litigation transfer — not present. The last ownership event precedes today by ~8 years and 2023-02-22 by ~4.3 years, and no infringement suit naming US 7,643,481 was found. A transfer that is not followed by a suit does not satisfy this signal.

6. Bankruptcy fire-sale — not present. Broadcom Corporation was acquired in the February 1, 2016 merger with Avago; it was not a Chapter 7 or Chapter 11 debtor, and the Broadcom/Avago successor did not file. The Bank of America entries are a consensual, secured financing lien (2016-02-11) later released (2017-02-03), not a bankruptcy or insolvency proceeding, and no sale-in-bankruptcy record exists for this patent.

7. Privateering — not present. Privateering requires a transfer from an operating company to an NPE that asserts on the operating company's behalf. Here the patent stayed inside the operating company's own group, and I found no assertion at all — no SEC disclosure, no Patent Progress/EFF coverage, no RPX/Unified listing.

8. Defensive aggregator (anti-NPE) — not present. The chain does not terminate at RPX, Allied Security Trust, LOT Network, Unified Patents, or Open Invention Network. The chain terminates at Avago Technologies International Sales Pte. Limited, a Broadcom Inc. group entity. The inverse observation applies instead: the patent was neutralized by attrition, not by acquisition — adjusted expiration 2023-02-22 for unpaid maintenance fees.

---

Verdict

Defensive / non-asserting.

The chain runs Broadcom Corporation → Avago Technologies General IP (Singapore) Pte. Ltd. (recorded 2017-02-01) → Avago Technologies International Sales Pte. Limited (merger recorded 2018-10-04, effective date corrected 2018-11-05, merger reel 047195/0827), with a Bank of America collateral lien recorded 2016-02-11 and released 2017-02-03 — five recorded events, all intra-group or lien-related, and zero NPE indicators. The patent has since lapsed for non-payment of maintenance fees (2023-02-22), so it is neutralized in fact, not merely unasserted. Definitional caveat: the rubric's bucket for this is written as "chain terminates at a defensive aggregator," and Broadcom is not RPX/AST/LOT/Unified — so this verdict is the closest fit rather than a literal match. The label that would be the literal match is "operating-company assertion," but that requires an actual assertion of this patent and I found none.

Known gaps, stated rather than papered over: (a) all reel/frame numbers except 047195/0827; (b) execution dates and the correspondent of record on every entry; (c) the inventor-to-Broadcom assignment records for the 1999 parents 09/343,409 and 09/527,999; (d) the granted claim text, which remains unretrieved from the prior section and is independent of this ownership analysis.

Verify at: https://assignment.uspto.gov/patent/index.html (search patent number 7643481) and https://assignmentcenter.uspto.gov/ — then open reel 047195/frame 0827 and the two 2017 records to capture the correspondent fields this session could not reach.

I'll research the prior art cited against US 7,643,481, starting with the patent's own citation records.

Let me pull the complete cited-references list and try to obtain the granted claims.

Let me try to obtain the claim text and the complete references list.

Prior Art Analysis — U.S. Patent No. 7,643,481 B2

"Network switch having a programmable counter" · App. No. 11/065,562 · Filed 2005-02-25 · Granted 2010-01-05 · Inventors Kadambi & Ambe · Original assignee Broadcom Corp.

---

0. Two threshold caveats you must read before the table

Caveat 1 — I still could not retrieve the granted claim text. Consistent with the earlier section of this analysis, the sources returned in this session (Google Patents, FreePatentsOnline, Justia, idiyas) gave me the bibliography, the abstract, the specification body, and the (56) reference lists — but not the granted claims. The FR text supplied to me is also truncated before the claims. A § 102 anticipation determination is a claim-limitation-by-claim-limitation exercise; without claim text I cannot truthfully write "Reference X anticipates claim N."

What I do instead below: I identify which cited references disclose the elements that the abstract and specification summary frame as the invention — (a) a source port, (b) a destination port containing the path to the destination, (c) a programmable counter unit counting a number of packets of selected packet types — and state, conditionally and explicitly flagged as an inference, which claim would be the one at risk if claim 1 tracks the abstract. All entries in the "potentially anticipatory claim" column are therefore conditional, not findings.

Caveat 2 — the § 102 date problem dominates this record. Under pre-AIA § 102 (this patent's claims were filed 2005-02-25, pre-AIAm so pre-AIA §§ 102/103 govern), the critical date is the earliest supported priority date — claimed as 1999-03-17 via provisional 60/124,878 → 09/527,999 (2000-03-17) → 11/065,562 (2005-02-25). The record also shows 11/065,562 was filed as a continuation-in-part, which matters enormously: any claim limitation not supported by the 1999/2000 parents gets only the 2005-02-25 date. The overwhelming majority of the cited references issued between 2001 and 2007. If the counter claims are entitled to the full 1999-03-17 priority date, most of the list below is not § 102 prior art at all and functions only as § 103/background art, or as art against the 2005 date. The examiner's reliance on this late-dated corpus is itself evidence about how the priority question was resolved in prosecution — but I could not retrieve the office actions to confirm which date was applied.

---

1. Tier 1 — the counter-specific references (highest § 102 relevance)

These are the two references the examiner cited that are directly about packet counters in a network device, and they are the only two in the list that squarely map to the "programmable counter unit … counting a number of packets" limitation.

Note on both: these are the two "counter" references the earlier section of this analysis flagged, and they remain the two most on-point. Neither, on the descriptions available to me, discloses programmable selection of packet types as the counting criterion — which is likely the distinction that carried the claims to allowance.

---

2. Tier 2 — packet classification and counter/action references (relevant to "selected packet types")

---

3. Tier 3 — switch-architecture references (elements, not the point of novelty)

Listed in the (56) record; each maps to pieces of the SOC 10 architecture described in the specification (slicing, mirroring, scheduling, stacking, memory management) rather than to the counter.

---

4. Tier 4 — the applicant's own earlier work (not § 102 prior art)

These appear in the (56) list but share the inventive entity (Kadambi et al.) and/or common ownership with the '481, and the '481 is itself a continuation of 6,996,099:

Under pre-AIA § 102(a)/(e), a reference is not "by another" where the inventive entity is the same or the subject matter was commonly owned at the time of invention. I flag this because a mechanical reading of the (56) list would wrongly treat the Kadambi family as anticipatory art.

---

5. Infrastructure/unrelated citations in the (56) list

Present for completeness; no § 102 exposure to the counter invention on the descriptions available: U.S. 6,912,223 (Sloane, automatic router configuration); U.S. 6,775,328 (Segaram, high-speed feedback synchronization loop); U.S. 6,640,707 — see Tier 2; U.S. 6,400,735 (Percey, glitchless delay line); U.S. 6,359,469 (Mendel, PLD logic element); U.S. 6,349,379 (Gibson, instruction flag for operand length); U.S. 6,336,180 (Long, virtual memory management); U.S. 6,330,584 (Joffe, multi-tasking resource sharing); U.S. 6,324,679 (Raghunathan, RTL power optimization); U.S. 6,324,279 (Kalmanek, two-phase signaling messages); US 2005/0094568 A1 (Judd, network path tracing).

---

6. Non-patent literature in the record

The Justia record shows NPL entries including the IBM Technical Disclosure Bulletin, "Local Area Network Switch Frame Lookup Technique for Increased Speed and Flexibility," 700 IBM TDB 38(7), July 1995, pp. 221–222, and Tanenbaum, Computer Networks, §§ 5.2–5.3, plus the Cisco Catalyst 8500 CSR Architecture white paper (1998). Caution: the same Justia excerpt intermixes Office Actions from unrelated applications (10/795,259; 11/976,862), so I cannot confirm which NPL entries belong to the 11/065,562 record versus a related case. The IBM TDB item (July 1995) and the Tanenbaum text are the ones that would predate 1999-03-17 and therefore retain § 102(b) status regardless of the priority resolution.

---

7. What I did not get, and what to pull next

1. The granted claims of 7,643,481 — required before any statement about "which claims are anticipated" can be upgraded from conditional to actual. Pull USPTO PatentCenter for 11/065,562, or the Google Patents "Claims" tab, or the US007643481 grant PDF.
2. The specification's counter disclosure — the FR text I hold is truncated mid-FFP description; the counter embodiment (and any claim-differentiating detail such as programming via CMIC 40 / programmable registers, or counting filtered packets via FFP 141) is beyond the truncation point. That disclosure is what determines whether Creedon '507 is overcome.
3. The full (56) foreign-reference list and the Office Actions — my FPO/Justia retrieval of the '481 reference list is partial (it truncates around the 2001-dated entries), and I could not retrieve the examiner's rejections. The rejections would tell us definitively which references the examiner applied under § 102 versus § 103, and against which effective date.
4. Full bibliographic verification of U.S. 5,610,905 (Murthy et al., 1997-03-11) — I confirmed the number and date only; I did not verify its title or disclosure in this session. Given it is the earliest-dated patent reference, it deserves a full read.

Bottom line: the most relevant prior art cited against US 7,643,481 is U.S. 5,987,507 (Creedon et al.) and U.S. 6,263,368 (Martin) — the two counter-specific references — with U.S. 6,343,130 (Lakshman) and U.S. 6,643,260 (Kloth) as the classification-side complements. I am declining to assign specific anticipatory claim numbers, because I could not retrieve the claim text and doing so would mean constructing claim scope rather than reading it.

I'll build on the prior sections, but first I need to close the gap they flagged: the claim text. Let me try again for the claims and for the prior-art references named in the record.

I have the Creedon reference confirmed. Now let me try harder for the actual claim text — including the parent US 6,996,099, which shares the title.

Obviousness Analysis — U.S. Patent No. 7,643,481 B2

"Network switch having a programmable counter"
Appl. No. 11/065,562 | Filed 2005-02-25 | Granted 2010-01-05 | Priority/prior-art date 1999-03-17
Inventors: Kadambi, Ambe | Assignee at issue: Broadcom Corp. | Now: Avago Technologies International Sales Pte. Ltd.
Governing law: pre-AIA 35 U.S.C. § 103(a) (effective filing date well before 2013-03-16)

---

0. Threshold caveat — read this before the analysis

This section builds on the previously generated material. That material flagged, correctly, that the granted claim text was not successfully retrieved. I attempted again this session (Google Patents, FreePatentsOnline, Justia, patentimages PDF, and phrase-searches on the abstract language). I still did not obtain the claim set. The full patent text supplied to me truncates before the claims.

Consequences for this §103 analysis, stated plainly:

• I can perform a claim-scope-conditional obviousness analysis. I cannot perform a limitation-by-limitation analysis, because I do not have the limitations.
• I will anchor the analysis on the only scope reliably established by the record: the abstract and the specification's own summary paragraph, which recite essentially identical language. Per the prior section, that is the strongest available signal about the head claim.
• Where a reference is characterized below, I mark whether I verified it this session or am relying on the earlier section.
• Do not treat the reconstructed claim language below as the actual claim. It is a working hypothesis only.

Reconstructed working hypothesis of claim 1 (UNVERIFIED — do not cite as claim text):

"A network switch for switching packets from a source to a destination, comprising: a source port for receiving an incoming packet from a source; a destination port that contains a path to a destination for said packet; and a programmable counter unit for counting a number of packets of selected packet types which are received by said switch."

If the granted claim instead recites specific structure (e.g., counter coupled to a specific module such as IPIC 90, a decrement path, a particular comparison against a watermark, or a specific programming path through CMIC 40), the analysis below must be re-run against that text. The absence of the claim text is the single largest uncertainty in this report, and I flag it rather than paper over it.

---

1. Legal framework

Under pre-AIA §103(a), the question is whether the subject matter as a whole would have been obvious to a person of ordinary skill in the art ("POSITA") at the time of invention. The Graham v. John Deere factors govern: (1) scope and content of the prior art; (2) differences between the prior art and the claims; (3) level of ordinary skill; and (4) objective indicia.

KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) controls the combination question:

• A combination of familiar elements according to known methods is obvious when it yields no more than predictable results.
• "[I]f a technique has been used to improve one device, and a person of ordinary skill in the art would recognize that it would improve similar devices in the same way, using the technique is obvious."
• A finite number of identified, predictable solutions makes the combination obvious.
• A design need or market pressure to solve a problem, plus a finite number of solutions, supports obviousness.
• "Obvious to try" suffices where the field presents a finite set of predictable options.

Applicant-admitted prior art in the specification counts as prior art. Here the specification itself recites that CPU 52 can access "MIB counters, programmable registers, status and control registers, configuration registers, ARL tables..." through CMIC 40. That is an express admission that programmable on-chip counters and programmable registers, CPU-accessible via a management interface, were known in this exact switch architecture.

Level of ordinary skill: a POSITA here is an engineer with a B.S.E.E./C.S. and roughly 2–4 years in network switch ASIC design, familiar with Ethernet switching, ARL/VLAN tables, shared-memory switch fabrics, and network management counters. The 1999–2000 timeframe is the relevant window; the art was mature.

Effective prior-art window: the '481 claims priority to 1999-03-17 (provisionals) and 2000-03-17 (Ser. No. 09/527,999). For claims first supported by 09/527,999, the window extends to 2000-03-17. Either way, references filed before 1999-03-17 are §102(b)/(e) art.

---

2. Prior-art landscape

Family references that are not prior art — exclude these

• US 6,996,099 ("Network switch having a programmable counter," Kadambi et al.) is the parent of the '481 in the same family chain (09/527,999). A parent sharing the same disclosure cannot be used as §102/§103 prior art against its own continuation.
• US 6,335,932 (from provisional-supporting Ser. No. 09/343,409) is likewise in-chain and unavailable.
• Any theory built on "the parent discloses it, so it's admitted art" is a legal error, not merely a weak theory. Flag it, do not use it.

---

3. The key inventiveness question, framed

The title and abstract reduce the alleged invention to three elements:

1. a source port receiving an incoming packet;
2. a destination port containing a path to the destination; and
3. a programmable counter unit for counting a number of packets of selected packet types received by the switch.

Elements (1) and (2) are the definition of a switch and are plainly old — they appear in every reference in §2 above and in the '481's own background discussion of "layer two switching-related devices… multiport bridges."

So the entire §103 question collapses to element (3): is a programmable, packet-type-selective counter in a multi-port switch obvious?

The prior art answers in the affirmative on three independent theories.

---

4. Combination A — Creedon '507 alone or in view of the applicant's own admissions

A.1 What Creedon teaches (verified).

Creedon discloses, in a device explicitly a "network device, such as a switch or repeater, which has a multiplicity of ports for the reception of data… in the form of addressed packets":

• A multiplicity of receive ports (2a, 2b, 2c) — source ports.
• A common buffer memory and read-out of data onward to the intended destination — a destination path.
• "Associated with each of the ports is a respective counter (8a, 8b, etc.) which may be cleared by the processor" — a counter, processor-controlled.
• The counters are enabled by the processing system on detection of a threshold: "If the common buffer memory approaches a congested state… these counters may be enabled to count."
• The counters increment as buffers are allocated and decrement as data is read out.
• "At the various watermarks, various actions can be taken, such as enabling flow control" — i.e., programmable thresholds with actions.
• "the counters may be cleared… If the memory becomes decongested at any stage, all the counters may be cleared."

A.2 Application to the presumed claim.

Creedon's per-port counters are programmable in the sense that the processor clears them, enables them, and the watermark values that trigger their operation and the resultant actions are programmable. Creedon is a multi-port switching device with source ports, destination paths, and processor-controlled counters. The '481's own specification confirms this architectural vocabulary was standard.

A.3 Where Creedon arguably falls short, and how that gap closes.

Creedon counts per port, not per packet type. That is the one genuine difference. It closes under §103 because:

• Selecting which packets to count is a known classification step. Lakshman '130 (packet classification by field) and the '481's own FFP 141 / rules table 22 disclose exactly this: parsing L2–L7 fields and taking a programmable action on match. A POSITA seeking per-type counting would simply gate a known counter with a known classifier.
• Motivation: network management (MIB/RMON counters), QoS policing, billing/accounting, and troubleshooting all require type-specific counts rather than raw per-port counts. The specification's own admission of "MIB counters" shows the demand existed.
• Reasonable expectation of success: the result is the predictable aggregation of two known, independently-operable techniques — a counter and a packet classifier. KSR: "familiar elements according to known methods… no more than predictable results." Nothing in Creedon teaches away; counting a subset of packets rather than all of them is a straightforward parameterization.

---

5. Combination B — Creedon '507 in view of the programmable counter array (US 5,687,173 / EP 0 917 783)

B.1 What the array-processor reference teaches.

This reference discloses, in a network switch, an "addressable high speed register array" holding a plurality of registers R₁…Rₙ for each supported connection, addressed via an index register loaded from a decoded packet header, with an operand register whose fields specify clear, set, increment, decrement, or retain, all updated in parallel. The reference expressly contemplates designating registers by data type — "one register may be reserved for counting the number of received cells for a particular connection… another for counting errors associated with the connection… another for counting cells with a cell loss priority of zero… and another for counting cells received at the switch which have a cell loss priority of one (1) or any other information which is deemed worthy of retention for purposes of network management."

B.2 Why this is the strongest combination.

This reference supplies the two features that jointly define element (3):

1. "Counter" — an array of counters that count received cells/frames in a switch.
2. "Programmable" — the register array is addressable and operand-driven; the operation (increment/decrement/clear/set) is selectable per update, and the type of thing counted is designated per register (received cells, error cells, cell-loss-priority-0 cells, cell-loss-priority-1 cells). Counting cells having a particular attribute is ipse facto counting selected packet types.

Combined with Creedon (multi-port switch, source ports, common buffer, destination read-out), the presumed claim reads on the combination directly.

B.3 Motivation to combine (KSR).

• Same field of endeavor: both are network switching devices with per-connection or per-port accounting hardware.
• Same problem: managing network traffic and accumulating operational statistics efficiently in real time.
• Predictable, advantageous result: the array processor's teaching states its own benefit — "large volumes of operational information regarding received cells/frames may be accumulated in real time… advantageously used… to efficiently manage the network traffic." That is an express, articulable motivation.
• Finite, predictable solution set: at the relevant time, there were only a handful of ways to count classified traffic (a register, a latch, a RAM-backed counter). KSR: obvious to try.
• Design incentive: the '481's own CMIC/MIB admission shows that a CPU-accessible counter bank was a recognized requirement in switch design.

---

6. Combination C — Creedon + Martin + the '481's admitted programmable-register architecture

C.1 The combination.

Creedon (multi-port switch with processor-controlled port counters) + Martin (counting message packets to/from a multi-computer server in a network for load-balancing purposes) + the applicant's admitted MIB/programmable-register/CMIC architecture.

C.2 Motivation.

Martin supplies the express purpose for counting packets: load balancing on a computing resource. A POSITA seeking to extend Creedon's port counters into a statistical/management function would look to Martin for a recognized application of packet counting in a switch. The applicant's own specification supplies the implementation vehicle: "CPU 52, through the CMIC 40, will be able to access numerous resources on SOC 10, including MIB counters, programmable registers, status and control registers, configuration registers…" A programmable counter bank reachable by a management CPU is a known architectural block, and adding selective counting is a use of that block for its intended purpose.

C.3 KSR point.

This is the paradigm "KSR predictable-use" case: each reference performs its own known function; the combination is the mere co-location of known counting, known classification, and a known programmable register interface. "[A] court must ask whether the improvement is more than the predictable use of prior art elements according to their established functions." Here it is not.

---

7. Secondary considerations (objective indicia)

On the record generated in the prior sections, no objective indicia support nonobviousness, and several cut against it:

• No commercial-success nexus was identified, and there is no evidence tying the '481's specific claims (as opposed to Broadcom's general switch business) to a product's success.
• No licensing or industry-acclaim evidence specific to this patent was found.
• No litigation: the prior section's search found no district-court, ITC, PTAB, or Federal Circuit proceeding naming the '481. A patent that was never asserted and lapsed for non-payment of maintenance fees (2023-02-22) supplies no evidence of the kind of industry-recognized breakthrough that rebuts a §103 prima facie case.
• No unexpected results are apparent from the disclosure: the counter operates as counters conventionally do (increment on receipt, decrement on transmission, compare against programmable watermarks) — the same mechanics Creedon describes.
• Patentability-by-narrow-claiming would not help here: if the granted claims were narrowed substantially beyond the abstract, that narrowness is an argument about claim scope, not about obviousness of the broader subject matter stated in the title and abstract.

---

8. Summary of the prima facie case

Bottom line: On the scope expressed in the '481's title and abstract, the subject matter would have been prima facie obvious under pre-AIA §103(a). The strongest formulation is:

US 5,987,507 (Creedon) in view of US 5,687,173 (network array processor / EP 0 917 783 B1), and further in view of a packet-classification reference such as US 6,341,130 (Lakshman).

The motivation is supplied by the references' own stated purposes (real-time accumulation of operational information for network management) and by the applicant's admitted MIB/programmable-register architecture. The result — selectively counting classified packets with a processor-programmable counter — is the predictable use of known elements according to their established functions.

---

9. Where the analysis is weakest — and what would change the answer

I will not overstate the case. Three caveats:

1. The claim text is unknown to me. If the granted claims recite, e.g., a counter specific to the IPIC 90 / network buffer pool 92 path; a counter that decrements in response to a CBM 71 decrement-cell-count message; or a counter whose count of selected packet types drives a specific FFP rules-table action, then a POSITA must be shown to have had reason to arrive at that particular configuration. Creedon and '173 support programmability and selective counting generally, but the specific wiring is where a validity challenge is most likely to founder — and where a defense would be built.

2. Creedon's enumerated purpose is congestion control, not statistics. A patent owner would argue Creedon's counters are enabled only on a congestion threshold, decremented on read-out, and cleared when decongested — i.e., transient flow-control instruments, not a general programmable accounting counter. The rebuttal runs through '173, whose express purpose is "statistical and operational information… for purposes of network management," and through the applicant's MIB admission. This is a rebuttable point, not a closed one.

3. The examiner already had Creedon and Martin of record (per the earlier section) and nonetheless allowed the case. That is weak evidence of nonobviousness — KSR cautions that "the grant of a patent… carries no presumption of validity that survives an obviousness challenge on different art or a different combination" — but it does suggest the examiner saw something in the claims that distinguished the cited combination. That "something" lives in the claims I could not retrieve, and it is the first thing I would pull from USPTO PatentCenter or the granted PDF (US007643481) before relying on this report.

Verification I still owe you (and could not complete in this session):

• Granted claim set: USPTO PatentCenter, Appl. 11/065,562; Google Patents "Claims" tab for US7643481B2; patentimages.storage.googleapis.com/.../US7643481.pdf.
• Prosecution history / IDS to confirm the actual prior-art references of record and any examiner reasons for allowance (any "Reasons for Allowance" or Notice of Allowance in the file wrapper).
• Confirmation of US 5,687,173's issue date, assignee, and exact title (I located the family but did not verify the bibliographic data).
• Confirmation of US 6,263,368 (Martin) filing date, to establish it as §102(e) art as of a date before 1999-03-17.

Until the claim text is in hand, treat the above as a robust framework and preliminary conclusion, not a final validity opinion.