US 6147976

I'll search for authoritative records on this patent number before summarizing.

I have solid data on the patent and no CAFC hit. Let me try to pin down the full claim set (the searches showed claims 1–17 but the patent reportedly has 19 claims).

US Patent 6,147,976 — "Fast Network Layer Packet Filter"

Search results note: I searched for the patent number as a literal string and for any CAFC 2026 docket referencing it. I found no CAFC 2026 docket, appeal, or IPR involving US 6,147,976. All hits for "6147976" were either the patent itself (Google Patents, Justia, SumoBrain, RPX, EveryPatent, FreePatentsOnline) or citations to it as prior art in unrelated later patents (e.g., US 7,295,561; US 7,272,625; US 10,411,734; US 2010/0014518). The "CAFC 2026" hits I retrieved concerned unrelated patents (e.g., BerkeleyIEOR v. W.W. Grainger*; the Globus Medical spinal-implant appeal). I state this as a negative search result, not as proof that no such proceeding exists — I do not have an authoritative CPAB/CAFC docket feed.

Bibliographic data (as recorded)

Assignment chain (per Google Patents reassignment records): Digital Equipment Corporation (inventor assignments, June 1996 and Oct. 1996) → Cabletron Systems, Inc. (Mar. 5, 1998) → Enterasys Networks, Inc. (Oct. 19, 2000) → Extreme Networks, Inc. (Aug. 26/Sept. 3, 2015), with intervening security-interest filings with Obsidian, Wells Fargo, and Silicon Valley Bank.

⚠️ Uncertainty flag: The Google Patents record contains an internal inconsistency — it lists both "Application filed by Cabletron Systems Inc" and "Assigned to DIGITAL EQUIPMENT CORPORATION" on the same 1996-06-24 date. I report both literally rather than reconciling them. Treat the original assignee designation as "as-recorded"; the enforceable chain is best confirmed from the USPTO Assignment database (assignment.uspto.gov).

Abstract (as published)

A packet filtering system associates domain identifiers with respective sets of addresses, where a set of domain prefixes defines the addresses belonging to a given domain identifier. Domain prefixes matching the packet's source and destination addresses yield source and destination domain identifiers. The system similarly associates protocol types plus port ranges with protocol domain identifiers; the packet's protocol type, source port, and destination port yield source and destination protocol domain identifiers. A filtering matrix holds entries each indicating whether a packet should be filtered. Source domain ID, destination domain ID, and the protocol domain IDs index into the matrix — e.g., as dimensions of a four-dimensional array. In an alternative embodiment, the source and destination protocol IDs obtain a protocol index from a protocol vector, and that protocol index plus the source and destination domain identifiers index a three-dimensional matrix. The matrix is further optimized for size, and lookup time depends only on the number of filtering domains, not on the complexity or number of filtering rules.

Plain-language overview of the independent claims

There are five independent claims: 1, 6, 16, 17, and 18. (Claims 2–5 depend from 1; 7–15 depend from 6; 19 depends from 18.)

Claim 1 — Core two-dimensional filtering system (broadest claim).
A system for filtering a received packet that: (a) has means for associating network station addresses with domain identifiers; (b) retrieves a source domain identifier from that association based on the packet's source address; (c) separately retrieves a destination domain identifier based on the packet's destination address; (d) includes a filtering matrix whose entries each contain a forwarding flag; (e) retrieves one matrix entry as a function of the source and destination domain identifiers; and (f) forwards the packet if the retrieved entry's forwarding flag holds a first value. Notably, claim 1 is address-only — it does not recite protocol or port matching.

Claim 6 — Extended system adding protocol and port-range matching.
A system that associates a first set of network-layer addresses with a first domain identifier and a second set with a second domain identifier; associates a network protocol combined with a range of source port numbers and a range of destination port numbers with a protocol index; includes a filtering matrix with forwarding flags; determines the packet's source address falls in the first set and destination address in the second set; uses protocol domain determining means to check that the packet's protocol type equals the network protocol and that its source port is within the source port range and its destination port is within the destination port range; and generates an index into the filtering matrix responsive to the first domain identifier, second domain identifier, and protocol index. This is the claim that captures the "filter on both source and destination TCP/UDP ports" contribution.

Claim 16 — System for building the data structure, with row-deduplication.
A system that, responsive to a command script, forms an address table (entries having a domain prefix and a domain identifier) and forms a two-dimensional filtering matrix with each row indexed by a domain identifier and each entry having a filtering flag. It then identifies a first matrix row that is identical to a second row (indexed by different domain identifiers) and deletes the second row, replacing the second domain identifier in the address table with the first. This claims the matrix-compression/optimization step at a structural level.

Claim 17 — System for building the data structure, with protocol-vector and plane-deduplication.
A system that forms a protocol table (protocol type, port range, associated protocol domain identifier), forms a two-dimensional protocol vector (entries indexed by a pair of protocol domain identifiers, each entry holding a protocol index), and forms a three-dimensional filtering matrix (first and second dimensions indexed by domain identifiers, third dimension indexed by a protocol index, each entry having a filtering flag). It identifies a first plane in the matrix identical to a second plane, deletes the second plane, and replaces the second protocol index in the protocol vector with the first. This claims the three-dimensional variant plus its plane-based optimization.

Claim 18 — Method counterpart to claim 1.
A method of filtering a packet received at a device: extracting a source address and a destination address from the packet; identifying source and destination domain identifiers associated with those addresses respectively; retrieving one entry from a plurality of entries in a filtering matrix as a function of the source and destination domain identifiers, each entry comprising a forwarding flag; and forwarding the packet if the forwarding flag in the retrieved entry equals a first value. (Claim 19, dependent, adds that the domain identifier indicates a network protocol associated with the address.)

Analytical notes

• Claim-style caveat: Every independent claim except 18 uses "means for" language. Given the June 1996 filing date (pre-Williamson), these are genuine means-plus-function recitations under 35 U.S.C. § 112 ¶ 6, and their scope would be limited to the corresponding structures disclosed (address table, protocol table, protocol vector, filtering matrix, Matrix Forming Code / Packet Filtering Code) and equivalents.
• Expiration: The patent expired June 24, 2016. Any residual infringement exposure would be limited to past damages within the statutory lookback period relative to a timely-filed suit; there is no possibility of ongoing injunctive relief.
• No article of manufacture or CRM claims: The claims are confined to systems and methods; there are no computer-readable-medium or "program product" claims.
• What the claims do not cover: Claim 1, the broadest, omits protocol/port filtering entirely (that appears first in claim 6). The inverse-generation technique, subdomain lists, "not" rules, reflection optimization (negative protocol vector values), and the intermediate-vector dimension-combining embodiment of FIG. 10 are described in the specification but I found no claim expressly directed to those features — they are disclosed but appear to be unclaimed subject matter.

Bottom line on your two specific requests: (1) The bibliographic and claim data above is well-corroborated across Google Patents, Justia, RPX, SumoBrain, and EveryPatent, which agree on all key fields. (2) I found no CAFC 2026 docket activity for this patent number, and given its 2016 expiration, appellate litigation in 2026 is unlikely — but I flag this as an unverified negative rather than a certainty.

I'll search multiple litigation-focused sources for this patent number.

Let me search litigation-specific databases with more targeted queries.

Bottom line

I found no litigation — no district court action, no CAFC appeal, no ITC investigation, and no PTAB proceeding — in which U.S. Patent No. 6,147,976 was the asserted (or challenged) patent. Every hit for the number "6147976" / "6,147,976" that I could retrieve resolves to one of three categories, none of which is an assertion of this patent:

1. The patent's own bibliographic record (Google Patents, Justia, FreePatentsOnline, SumoBrain, Unified Patents' patent-profile page).
2. Citations of 6,147,976 as prior art in other patents and prosecution papers (see detail below).
3. Unrelated documents where "6147976" is a trademark registration number (Willow Industries, KADANCE) or a Japanese patent number (JP6147976B2, an LED package) — i.e., false-positive string matches, not US 6,147,976.

I must be explicit about the limits of this: I was cut off before I could query the Unified Patents litigation caselist, RPX Insight, Docket Alarm's case database, or PACER directly. So this is a strong negative search result, not a certified absence. Given the patent expired June 24, 2016 and is 24+ years from issue, the absence is also plausible on its face.

Cases that superficially match — and why each is NOT this patent

These are the near-misses I want expressly distinguished, because they are the most likely sources of an erroneous "there was litigation" conclusion.

Note on the AT&T-style trap here: "the '797 patent" is used loosely across unrelated litigations. None of those is 6,147,976, whose own short form would be "the '976 patent."

The only substantive third-party "hits" — all prior-art citations, not litigation

What I checked, and where the gaps are

Checked (via search): Google Patents record and citation lists; Justia; FreePatentsOnline; SumoBrain; Unified Patents patent-profile pages; Docket Alarm document corpus; CourtListener; cases.justia.com; EPO search-report annexes; news of Extreme/Enterasys litigation.

Not checked before I was cut off (recommend verifying):

• https://portal.unifiedpatents.com/litigation/caselist filtered to patent 6147976
• PTAB (IPR/PGR/CBM) search for the patent as challenged patent
• PACER / CourtListener RECAP docket search on "6,147,976" and "6147976" as asserted patent
• USPTO Patent Center / Global Dossier "Litigation" tab (which surfaces any district-court records the Office has indexed)
• ITC (Section 337) EDIS search
• District of Delaware / E.D. Tex. dockets, where Enterasys/Cabletron-era networking patents were most often enforced

If you want a definitive answer, run the exact-string docket queries "6,147,976" and "6147976" on Unified Patents' litigation search and Docket Alarm, and optionally a PACER Case Locator search restricted to patent nature-of-suit codes. I can also run a LexMachina/Intelligize-style query if you have access.

Practical read for a client

• No assertion of 6,147,976 is known to me. The patent expired June 24, 2016 (anticipated expiration, per the Google Patents legal-status record), so there is no injunctive exposure and no ongoing-royalty theory; any residual claim would be a past-damages theory subject to the § 286 six-year lookback as measured from a timely-filed complaint.
• The only widely recorded use of the patent is defensive — as "X" prior art in an EPO search report (EP 1 517 518) and as a cited reference in later U.S. patents. That is the profile of a patent cited in prosecution, not litigated.
• ⚠️ Do not attribute the Extreme Networks v. Enterasys cases to this patent. They involved the same corporations but different patents ('727, '181). If a source tells you otherwise, treat it as a number-mangling error.

Caveat restated: absence of evidence in my searches is not proof of absence. Because I could not complete queries against the primary litigation databases, I would not represent to a court or a client that "no suit ever named 6,147,976" on the strength of this search alone.

I'll verify the structured "no proceedings" result against public sources before writing this up.

Let me run a few more targeted checks before concluding.

One more verification pass on PTAB-specific indexes.

Proceedings overview

Total AIA trial proceedings on file for US 6,147,976: 0. The structured USPTO Open Data Portal (ODP) block in this prompt — the canonical source — returns no IPR, no PGR, and no CBM naming this patent as the challenged patent. Breakdown by status is therefore all zeros: 0 active, 0 with claims invalidated, 0 with claims sustained, 0 settled, 0 institution denials. My independent web searches (Google Patents/Justia/FPO citation tables, Docket Alarm PTAB document corpus, PTAB petition PDFs, CAFC opinion text, and third-party PTAB commentary) surfaced no proceeding number of any kind in which 6,147,976 is the patent under review.

Bottom-line defensive posture: there is no IPR-canceled claim to hand a defendant, and equally no adverse Final Written Decision and no § 315(e)(2) estoppel to worry about. Every one of the five independent claims (1, 6, 16, 17, 18) and all 14 dependent claims is untested at the Board. But this absence is not a hardening signal — it is an artifact of the calendar. The patent issued 2000-11-14, AIA trials did not exist until 2012-09-16, the patent expired 2016-06-24, and (per the litigation section of this analysis) no district court assertion of it is known. IPRs are filed by defendants; with no assertion, there is no petitioner. If a demand letter today cites claims 1 or 6, the correct response is not an IPR — it is "this patent expired 2016-06-24 and I have no record of any live case" (see Recommended next steps).

Proceeding-level detail

None exists. I am not going to populate the per-proceeding template with invented docket numbers, panels, or dispositions. There is no FWD to quote, no institution decision to characterize, and no appeal to trace.

Two consequences worth stating precisely:

• No § 315(e)(2) estoppel is in force against anyone. Because no IPR reached a final written decision, no petitioner and no privy is barred from raising any § 102/§ 103 ground in district court. Conversely, a defendant gets no "PTAB already found this invalid" shortcut.
• No Arthrex-style or Director-review activity exists on this patent either — that would be downstream of an FWD that never issued.

Why the ODP result is plausible on its face (not just a data gap)

• Temporal window. AIA trials opened 2012-09-16; the patent expired 2016-06-24 (anticipated, per the Google Patents legal-status record). A party wanting to challenge it had a ~3 years 9 months window during the enforceable, post-AIA term. That is short for an unasserted patent.
• CBM was never a fit. Covered Business Method review requires a pre-AIA, financial-services-related patent and sunset on 2020-09-16. This is a 1996 network-layer packet-filter patent; no CBM pathway was realistically open.
• Assertion is the precondition. ~87% of IPR petitioners are district court defendants (per the practitioner data surfaced in my searches). The litigation section of this analysis found no assertion — so the expected IPR population is zero.
• The patent's actual footprint is defensive. It shows up as "X" (particularly relevant) art against claims 1–17 in EPO search report EP 1 517 518, and as a cited reference in dozens of later U.S. patents. Citee, not challengee.

False positives to avoid (do not import these into a client memo)

The searches returned several nearby hits that a hurried analyst would miscaption as PTAB proceedings on this patent. Each is wrong, for a specific reason:

Rule for the file: any citation to "6147976" that appears inside an exhibit list, an IDS, a "References Cited" table, or an EP/JP/WO search report is a citation, not a proceeding. Demand a proceeding number in the form IPR20XX-XXXXX, PGR20XX-XXXXX, or CBM20XX-XXXXX with US 6,147,976 as the challenged patent.

Strategic summary

Claim status across the whole patent: UNTESTED. No claim of 6,147,976 has ever been canceled, confirmed, or even institution-tested by the PTAB. Independent claims 1 (address-only filtering system), 6 (system adding protocol + source/destination port-range matching), 16 (system for forming the data structure with row deduplication), 17 (system with protocol vector and three-dimensional matrix with plane deduplication), and 18 (method counterpart to claim 1) all stand as issued. Dependents 2–5, 7–15, and 19 likewise. This is the mirror image of the "cancelled claims" scenario — there is nothing for a defendant to point at and nothing for a patent owner to hide behind.

Estoppel landscape: empty. With no FWD, § 315(e)(2) estoppel never attached to any petitioner or privy. A defendant being asserted today (if such assertion exists) may raise any § 102 or § 103 ground based on patents and printed publications in district court, including grounds a hypothetical petitioner "reasonably could have raised." There is no prior PTAB record to fight over, no claim construction from the Board to reconcile, and no IPR to stay a case for. The trade-off is symmetrical: the patent owner also faces no adverse estoppel or invalidity finding it would have to distinguish — but with a 2016-06-24 expiration, that distinction is academic.

Pattern signals: none of the usual ones. No serial petitioner (no petitioner at all). No patent owner PTAB appeal practice on this patent — consistent with the litigation section's finding that the known Extreme Networks v. Enterasys activity involved different patents (the '727 and '181), not 6,147,976. No defensive aggregator is in this patent's PTAB chain; the only Unified Patents appearance is a patent-profile page listing 6,147,976 in a citations list for a different patent (US-11012415-B2), which is the same citation-not-challenge artifact flagged above. (Note the minor inconsistency the litigation section already flagged: that profile shows priority 1996-06-23, one day off the recorded filing date — reported literally, not corrected.)

Recommended next steps

If you are a defendant facing an assertion of US 6,147,976:

1. Start with the expiration, not the PTAB. The patent's anticipated expiration is 2016-06-24 (Google Patents legal-status record, "Expired – Lifetime"). There is no possibility of injunctive relief and no ongoing-royalty theory; any exposure is past damages bounded by the § 286 six-year lookback measured from a timely-filed complaint. Confirm at https://patents.google.com/patent/US6147976/en and against USPTO Patent Center maintenance-fee records.
2. There is no PTAB proceeding to cite — so do not promise your client one. I am stating this as a strong negative from the ODP block plus targeted searching, not as a certificate of absence. Before filing anything, run the exact-string searches against the primary sources yourself:
   • PTAB E2E / PTAB API: https://ptacts.uspto.gov/ptabweb/ and https://developer.uspto.gov/ptab-api/ — query the patent number as the challenged patent, not as a cited reference.
   • Unified Patents litigation caselist: https://portal.unifiedpatents.com/litigation/caselist, filtered to 6147976.
   • CourtListener / RECAP string search: "6,147,976" and "6147976" at https://www.courtlistener.com/ — this catches any CAFC appeal that would exist only if an FWD had issued.
   • USPTO Patent Center → Global Dossier → "Litigation" tab, which surfaces any district-court records the Office has indexed.
3. Do not file an IPR defensively. With the patent expired, an IPR buys no stay leverage and no invalidity judgment of practical value, and it forfeits nothing the defendant currently has. If a district court case is nonetheless live, the cheaper path is a § 282 invalidity defense plus the expiration/damages arguments — all grounds remain available precisely because no IPR estoppel attached.
4. If anyone represents that this patent "has claims canceled in IPR," demand the proceeding number in IPR20XX-XXXXX form with US 6,147,976 as the challenged patent, and cross-check it against PTAB E2E. On the record before me, no such number exists.

Confidence statement. The ODP block is authoritative and returns zero proceedings; my web searches are consistent with it and uncovered no contrary proceeding. I was unable to complete independent queries against the PTAB API, Unified Patents caselist, and PACER/CourtListener directly (tool-call limit reached), so I characterize this as a strong negative search result, well-corroborated across two independent source types, but not a certified absence. The correlated absence of district court assertion is a strong tell that the PTAB absence is real rather than a data gap.

I'll search for the assignment record details, including reel/frame numbers and recording correspondents, before assembling the chain.

Let me look for the assignment recording details and any correspondent-of-record or NPE signals.

Let me try to retrieve the specific assignment records (reel/frame and correspondents) from patent-tracker and assignment mirrors.

Assignment & Ownership Reconstruction — US 6,147,976

Important scope note up front: I could not retrieve the USPTO Patent Assignment Center abstract of title for this patent. Assignment Center (assignmentcenter.uspto.gov) and its legacy mirror (assignment.uspto.gov/patent/index.html) are JavaScript single-page applications whose record pulls did not surface via web search, and I exhausted my search budget before reaching them. I therefore have no reel/frame numbers for the ordinary conveyances, and no correspondent-of-record (recording attorney/agent) data at all. I will not invent either. Everything below is drawn from (a) the Google Patents reassignment record in the authoritative full text, (b) reel/frame references that are cited inside two release documents on that record, and (c) independent corroboration from SEC filings and news for the corporate transactions. Where a field is unavailable, I say so rather than filling it.

Inventors

Basis for the employer call: the application was filed 1996-06-24 and all three inventors executed assignments to DIGITAL EQUIPMENT CORPORATION — Shand and Welch on a record dated 1996-06-24, Harper on 1996-10-11 (both per Google Patents legal events). A filed application assigned to DEC at filing is presumptively an employer assignment. The UK/French residences are consistent with DEC's European network-products engineering footprint (Surrey/Hampshire in the UK; Sophia Antipolis/Valbonne in France), which is the same business unit that was later sold on — see next section.

Unusual pattern — flagged

The original assignee's entire business unit was sold out from under the patent ~19 months after filing. DEC's network products business ("NPB") was acquired by Cabletron Systems, Inc. and its subsidiary Ctron Acquisition, Inc. on 1998-02-07, under an Asset Purchase Agreement dated 1997-11-24 (confirmed by Cabletron's Form 8-K, Item 2, date of earliest event reported 1998-02-07). This is not the "all inventors depart within 12 months" pattern — it is a corporate divestiture of the whole unit, which achieves the same result: the inventors' work product moved to an unrelated acquirer while the inventors themselves had no say in it. The causal chain DEC → Cabletron → Enterasys → Extreme is therefore a chain of asset and company sales, not inventor mobility.

⚠️ I did not verify whether any of the three inventors personally moved to Cabletron, Enterasys, or Extreme. I state that as unchecked, not as a negative.

Original assignee

Original assignee on the issued face: Cabletron Systems, Inc. (per Google Patents "Original Assignee" field).

• Primary line of business: Cabletron was a US (Rochester, New Hampshire) enterprise networking equipment vendor — LAN switches, hubs, routers, network management. It shipped products in the same technical space as the claims (screening routers / packet filters), so the patent was plausibly embodied in Cabletron/Enterasys router firmware. ⚠️ I have no documentary evidence that a specific Cabletron or Enterasys product implemented the claimed filtering matrix — the specification's FIG. 5 "Screening Router" is a generic embodiment. Treat "shipped a product embodying the claims" as plausible but unverified.
• Status: Cabletron Systems, Inc. no longer exists as such. It was renamed/reorganized and its enterprise networking business was spun off as Enterasys Networks, Inc. in 2000 (see below); the Cabletron name was later used for a residual entity.
• Current status of the patent's owner: Extreme Networks, Inc. is an operating, NASDAQ-listed Ethernet switching vendor (Santa Clara, California). The patent itself is Expired – Lifetime, anticipated expiration 2016-06-24, per the Google Patents legal-status record.

⚠️ Contradiction with the face of the patent — re-flagged and reconciled

The previously generated section flagged that Google Patents shows both "Application filed by Cabletron Systems Inc" and "Assigned to DIGITAL EQUIPMENT CORPORATION" on 1996-06-24. Reconciling with the assignment evidence: Cabletron had no relationship to this application in 1996. Cabletron appears on the record only as the 1998 assignee from DEC, and it is listed as "Original Assignee" on the issued face solely because it owned the application on the 2000-11-14 issue date. The "filed by Cabletron" attribute on Google Patents is a downstream artifact, not a 1996 filing event. The real 1996-06-24 event is the inventor-to-DEC assignment. This resolves rather than repeats the prior section's inconsistency flag.

Assignment timeline

Method note on dates and reel/frames: Google Patents legal events do not distinguish execution date from recordation date; the dates below are as recorded by Google Patents. For inventor assignments I believe they are execution dates (Shand/Welch on the filing date; Harper ~3.5 months later). For the corporate assignments I cannot tell. Reel/frame numbers are unavailable except where a release document itself quotes them — I reproduce those two literally.

1. 1996-06-24 (as recorded) / recordation date not exposed — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Assignors' Interest (inventor-to-employer)
• Assignor: Shand, Ian Michael Charles; Welch, Steven Richard
• Assignee: Digital Equipment Corporation
• Correspondent: not retrieved — see gap note below
• Context: original inventor assignment to employer at filing.

2. 1996-10-11 (as recorded) / recordation date not exposed — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Assignors' Interest
• Assignor: Harper, John Anthony
• Assignee: Digital Equipment Corporation
• Correspondent: not retrieved
• Context: original inventor assignment — filed separately and ~3.5 months after the other two. Minor oddity: the third inventor's paper was recorded separately and months later; I cannot explain this from available records (could be an administrative re-do, a foreign-residence paperwork delay, or a correction). Not evidence of anything.

3. 1998-03-05 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Assignors' Interest
• Assignor: Digital Equipment Corporation
• Assignee: Cabletron Systems, Inc.
• Correspondent: not retrieved
• Context: asset sale / divestiture. DEC's network products business was sold to Cabletron and Ctron Acquisition, Inc. under an APA dated 1997-11-24, closing 1998-02-07; this recording on 1998-03-05 perfects Cabletron's title to the portfolio. This is the only step by which Cabletron became "original assignee."

4. 2000-10-19 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Assignors' Interest
• Assignor: Cabletron Systems, Inc.
• Assignee: Enterasys Networks, Inc.
• Correspondent: not retrieved
• Context: internal reorganization / spin-off. Cabletron reorganized in 2000 into separate companies — Enterasys Networks, Riverstone Networks, Aprisma Management Technologies and Global Network Technology Services — with the enterprise networking business going to Enterasys. The October 2000 recording is the papering of that separation. ⚠️ I am confident on the substance of the Cabletron four-way split; I could not re-verify the exact spin-off effective date this session.

5. 2006-05-23 (as recorded) — Reel 017656 / Frame 0552 (inferred; see note) — Reel/Frame as quoted elsewhere in this file: 17656/0552

• Conveyance: Security Agreement (lien, not a title transfer)
• Assignor: Enterasys Networks, Inc.
• Assignee: Obsidian, LLC; Wells Fargo Foothill, Inc. (as lenders/agents)
• Correspondent: not retrieved
• Context: securitization. Patent portfolio pledged as collateral. Consistent with the 2006 take-private of Enterasys by Enterprise Networks Holdings, Inc. (the entity that, per Extreme's 2013 Form 8-K recitals, was the direct and beneficial owner of 100% of Enterasys' stock, with Enterprise Networks Holdings, B.V. as parent and guarantor). ⚠️ The reel/frame 17656/0552 is quoted inside the 2010-11-30 release entry as the security being released; I am inferring it corresponds to this 2006-05-23 lien. Flagged as inference.

6. 2010-11-10 (as recorded) — Reel 025339 / Frame 0875 (inferred) — Reel/Frame as quoted elsewhere in this file: 25339/0875

• Conveyance: Grant of Security Interest in U.S. Patents
• Assignor: Enterasys Networks, Inc.
• Assignee: Wells Fargo Trust Corporation Limited, as Security Agent
• Correspondent: not retrieved
• Context: securitization / refinancing. Replaced the 2006 Obsidian/Foothill facility. ⚠️ Reel/frame inferred from the 2013-11-01 termination entry which quotes 25339/0875.

7. 2010-11-30 (as recorded) — Reel 017656 / Frame 0552 (as expressly quoted in the entry)

• Conveyance: Release and Reassignment of Patents and Patent Applications at Reel/Frame No. 17656/0552
• Assignor: Enterprise Communications Funding GmbH (as successor in interest to Obsidian, LLC); Wells Fargo Capital Finance, Inc. (formerly Wells Fargo Foothill, Inc.)
• Assignee: Enterasys Networks, Inc.
• Correspondent: not retrieved
• Context: lien release following payoff/refinance — not a title transfer. This is the one entry where the record gives a hard, citable reel/frame.

8. 2013-11-01 (as recorded) — Reel 025339 / Frame 0875 (as expressly quoted in the entry)

• Conveyance: Termination and Release of Security Interest in Patents at Reel/Frame No. 25339/0875
• Assignor: Wells Fargo Trust Corporation Limited
• Assignee: Enterasys Networks, Inc.
• Correspondent: not retrieved
• Context: lien release. Timing is significant: Extreme Networks, Inc. closed its acquisition of Enterasys on 2013-10-31 for $180 million cash under a Stock Purchase Agreement dated 2013-09-12 between Extreme and Enterprise Networks Holdings, Inc. The secured party released its lien the day after closing.

9. 2015-07-27 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Security Agreement
• Assignor: Enterasys Networks, Inc.
• Assignee: Silicon Valley Bank
• Correspondent: not retrieved
• Context: securitization. ⚠️ Ordering oddity: if Entersys had already conveyed title to Extreme in the 2013 acquisition, a 2015 pledge by Enterasys as assignor is hard to square — unless (a) the Extreme assignment papers were not yet executed (see #10–11, recorded a month later), or (b) Enterasys remained an Extreme subsidiary pledging as co-obligor under the SVB facility. I lean to (b) + de facto control, but flag the sequence as not fully explained by the records I hold.

10. 2015-08-26 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Interest
• Assignor: Enterasys Networks, Inc.
• Assignee: Extreme Networks, Inc.
• Correspondent: not retrieved
• Context: transfer-to-acquirer. Recorded ~22 months after the 2013-10-31 deal closing — a late perfecting recording of an acquisition, not a fresh transaction.

11. 2015-09-03 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Assignment of Interest
• Assignor: Enterasys Networks, Inc.
• Assignee: Extreme Networks, Inc.
• Correspondent: not retrieved
• Context: duplicate / confirmatory recording. Two same-parties assignments eight days apart from the same assignor. Most likely one recording for a batch of US patents and a second for a different batch or a correction. Not a cascading NPE transfer.

12. 2016-06-24 — Not an assignment. Anticipated expiration (legal status).

13. 2016-10-31 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Amended and Restated Patent and Trademark Security Agreement
• Assignor: Extreme Networks, Inc.
• Assignee: Silicon Valley Bank
• Correspondent: not retrieved
• Context: securitization / refinancing. Post-expiration recordation; the security instrument blankets the portfolio regardless of this patent's lapse.

14. 2017-07-14 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Second Amended and Restated Patent and Trademark Security Agreement
• Assignor: Extreme Networks, Inc.
• Assignee: Silicon Valley Bank
• Correspondent: not retrieved
• Context: securitization / refinancing.

15. 2018-05-01 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Release by Secured Party
• Assignor: Silicon Valley Bank
• Assignee: Enterasys Networks, Inc.
• Correspondent: not retrieved
• Context: lien release. Paired with #16.

16. 2018-05-01 (as recorded) — Reel [not retrieved] / Frame [not retrieved]

• Conveyance: Release by Secured Party
• Assignor: Silicon Valley Bank
• Assignee: Extreme Networks, Inc.
• Correspondent: not retrieved
• Context: lien release. The two parallel releases on the same day confirm Enterasys was still a pledging party alongside Extreme in the SVB facility — which supports the reading in #9.

Correspondent-of-record gap — the single most important missing datum

The task correctly identifies the correspondent as the highest-value field for detecting NPE families. I retrieved zero correspondent names. I will not guess. Practically, the records to pull manually are the sixteen conveyances above from the Assignment Center abstract of title for patent 6,147,976 (search → "Abstract of Title"), which exposes the recording attorney/firm for each. Two of the reel/frames are already known and can be pulled directly: 017656/0552 and 025339/0875.

Timeline diagram

timeline
    title Ownership of US 6147976
    1996 : Filed 24 Jun by DEC inventors
         : Inventors assign rights to Digital Equipment
    1998 : DEC network business sold to Cabletron
    2000 : Enterasys takes rights from Cabletron
         : Patent issued 14 Nov
    2006 : Enterasys pledges portfolio to lenders
    2010 : Enterasys pledges to Wells Fargo Trust
    2013 : Extreme buys Enterasys
    2015 : Rights recorded to Extreme Networks
    2016 : Patent expires 24 Jun
    2018 : Lender releases recorded

NPE / troll-pattern signals

1. Shell-entity transfer — Not present

Every title-transferring assignee in the chain is a known operating networking vendor: Digital Equipment Corporation → Cabletron Systems, Inc. (1998-03-05) → Enterasys Networks, Inc. (2000-10-19) → Extreme Networks, Inc. (2015-08-26, 2015-09-03). There is no entity with an "IP / Patents / Licensing / Holdings / Ventures" suffix anywhere in the chain, no single-purpose LLC, and no registered-agent-service address. The only non-operating names in the record — Obsidian, LLC, Wells Fargo Foothill, Enterprise Communications Funding GmbH, Wells Fargo Trust Corporation Limited, Silicon Valley Bank — are each acting as secured lender or releasing secured party, i.e. lienholders, not owners (conveyance types "Security Agreement," "Grant of Security Interest," "Release by Secured Party").

2. Known asserter in the chain — Not present

None of the assignees or lienholders matches a public NPE list (Acacia, Marathon, IV, IPNav, Wi-LAN, Conversant/Mosaid, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, Spangenberg entities, etc.). Extreme Networks is in fact the inverse: a frequent NPE defendant. Per Extreme's own SEC disclosures it was sued by Orckit IP, LLC (German actions, EP 1 958 364 B1 and EP 3 068 077 B1) and by Global Innovation Aggregators, LLC (six Chinese patents, January 2019), and it was a defendant alongside Enterasys in Network-1's Remote Power Patent campaign (U.S. 6,218,930). A company that spends its 10-Q ink defending NPE suits is not running one. Also note the earlier Extreme Networks, Inc. v. Enterasys Networks, Inc., Fed. Cir. 09-1325 (2010) — that appeal involved the '727 and '181 patents, not 6,147,976, and it predates Extreme's 2013 acquisition of Enterasys. Do not read it into this chain.

3. Repeat correspondent across the chain — Unclear (no data)

This signal cannot be evaluated. No correspondent names were retrieved for any of the sixteen recorded events. Recurrence cannot be assessed without them. This is the one signal I would most want filled before finalizing the verdict.

4. Cascading transfers — Not present (a cluster exists, but it is financing, not shell-cascading)

There is a dense cluster of recordations: 2006-05-23, 2010-11-10, 2010-11-30, 2013-11-01, 2015-07-27, 2015-08-26, 2015-09-03, 2016-10-31, 2017-07-14, 2018-05-01 (x2) — ten events in twelve years, and three within 2015 alone. But the conveyance types defeat the NPE reading: they are dominantly "Security Agreement," "Grant of Security Interest," "Amended and Restated…," and "Release by Secured Party." That is the footprint of a leveraged operating company repeatedly pledging and re-pledging a portfolio to lenders (the 2006 take-private of Enterasys, then the SVB facility after the 2013 Extreme acquisition), not of a patent flipped through chained shells. The two genuinely odd items are the 22-month delay between the 2013-10-31 closing and the 2015-08/09 recordings, and the two same-party assignments eight days apart — late/duplicate perfecting, not cascading.

5. Pre-litigation transfer — Not present

No infringement suit naming 6,147,976 was found in the previously generated litigation section, so there is no suit for a 2015 transfer to precede. The two title transfers into Extreme (2015-08-26, 2015-09-03) land ~19 months before the patent's 2016-06-24 expiration and are explained by the 2013 corporate acquisition, not by litigation staging.

6. Bankruptcy fire-sale — Not present

No Chapter 7 or Chapter 11 proceeding appears anywhere in this chain. Enterasys' trajectory was distress → take-private → strategic sale, not bankruptcy: the 2006 sale to Enterprise Networks Holdings, Inc. (per the recitals of Extreme's 2013-09-12 Stock Purchase Agreement, that entity held 100% of Enterasys' stock, with Enterprise Networks Holdings, B.V. as parent/guarantor) and then the 2013-10-31 cash sale of Enterasys to Extreme Networks for $180 million. This is an LBO-to-strategic exit, not a Kodak/Nortel-style asset liquidation. ⚠️ I could not verify this session whether Enterasys ever filed for bankruptcy abroad or whether any subsidiary did; I state the absence from this chain only.

7. Privateering — Not present

Privateering requires an operating company to hand patents to an NPE that then sues the operating company's competitors on its behalf. Here the operating company (Extreme) acquired the patent's owner outright and retains the patent itself; nothing in the record shows a transfer out to a separate asserting vehicle. There is no SEC 8-K, news item, or Patent Progress/EFF coverage in my results reporting Extreme as a privateer with respect to this patent.

8. Defensive aggregator (anti-NPE) — Not present

The chain terminates at Extreme Networks, Inc., an operating company — not at RPX, AST, LOT Network, Unified Patents, or OIN. The 2018-05-01 releases are from Silicon Valley Bank (a lender releasing collateral), which is the only entity that could be mistaken for an aggregator and is not one. Inverse signal accordingly does not apply.

Verdict

Defensive / non-asserting — with an explicit definitional caveat.

Caveat: the option list defines this category as "chain terminates at a defensive aggregator." That is not literally true here — the chain terminates at Extreme Networks, Inc., an operating Ethernet-switching vendor. I am selecting the non-asserting half of the category because the two alternatives are affirmatively excluded: there is no shell/licensing entity anywhere in the chain (signal 1, not present), no known asserter (signal 2, not present), and no known suit naming this patent (per the prior litigation section). The record is a straight operating-company line: DEC → Cabletron (1998-03-05) → Enterasys (2000-10-19) → Extreme (2015-08-26 / 2015-09-03), punctuated only by lender liens and releases (R/F 017656/0552; R/F 025339/0875; and the 2015/2016/2017 SVB security agreements released 2018-05-01). The patent expired 2016-06-24, so no assertion is commercially possible now.

Confidence caveat: my two strongest signals (recurrence of correspondent, and any late-chain transfer to an unknown party) are unresolved because I retrieved no correspondent data and no reel/frame for the ordinary conveyances. If you want this verdict hardened, the single highest-value next step is to pull the Abstract of Title for patent 6,147,976 from the Assignment Center and, for each of the sixteen events, record the correspondent attorney/firm. A single name recurring across the 1998, 2000 and 2015 transfers would materially change the signal-3 analysis; its absence would confirm the reading above.

Verify at:

• https://assignmentcenter.uspto.gov/ (search by patent number 6147976 → Abstract of Title)
• https://assignment.uspto.gov/patent/index.html (legacy mirror, same underlying records)
• Direct reel/frame lookups: 017656/0552 and 025339/0875

I'll research the prior-art citations for this specific patent number. Let me run targeted searches.

Let me get the front-page "References Cited" list for the patent itself (not the patents that cite it).

Let me try to retrieve the patent's own front-page citation list directly.

I hit my tool-call ceiling mid-research, so let me be precise about what I actually retrieved versus what I could not — this matters a great deal for a §102 analysis, because the direction of a citation changes its legal relevance completely.

1. Critical retrieval gap — read this first

I could not retrieve the front-page "(56) References Cited" list of US 6,147,976 itself. Every query I ran returned one of two things instead:

• Forward citations — patents that cite 6,147,976 (Google Patents' "Patent Citations"/"Cited By" tables belonging to other patents, in which US6147976A appears as a cited reference), and
• Third-party uses of 6,147,976 as prior art against other patents (the EPO search report, examiner citations in later cases).

Google Patents renders the patent's own cited-art list under the heading "Patent Citations" on the US6147976A page. My fetches of that page returned the specification, claims, classifications, and abstracts — not the (56) block. The authoritative patent text supplied for this task likewise contains only the specification/claims; it has no "References Cited" front page, and the specification body cites no patents (only functional works: Backus-Naur Form, 4.2BSD UNIX reserved ports, the X Window System "Xserver" ports, and the "Longest Prefix Matching" technique).

Consequently: I cannot enumerate "each patent citation for 6147976" from retrieved evidence, and I will not fabricate that list. What I can give you is (a) every well-grounded citation relationship I actually found, and (b) a clearly-labeled analytical §102 candidate set. Treat (b) as a research worklist, not as the patent's recorded citation list.

2. The one confirmed citation relationship that is about prior art

Direction warning: this is 6,147,976 being used as "X" (particularly relevant) prior art against EP 1 517 518's claims 1–17 — not prior art against 6,147,976. It is the single most legally meaningful third-party citation to this patent that I found, and it is on the offensive side.

3. Forward citations — patents citing 6,147,976 (explicitly NOT its prior art)

These appear in the "citations" tables of other patents. Because they post-date the June 24, 1996 filing, they cannot anticipate 6,147,976; they are evidence of the field's evolution only.

Also note: US 5,951,651 ("Packet filter system using BITMAP vector of filter rules," Lucent, priority 1997-07-23) and US 5,921,090 / US 6,141,749 appear in these tables too — but those are citations of the host patent, and in any event several have post-1996 priority, so they are not §102 art against 6,147,976.

4. Candidate §102 prior art — analytical, NOT a retrieved citation list

Legal frame (pre-AIA applies; caucus date = June 24, 1996): §102(b) art requires publication/patenting before June 24, 1995; §102(a) art requires public knowledge/use/patenting before the invention date (presumptively just before the 1996 filing); §102(e) art requires a U.S. patent or published application with an effective filing date before June 24, 1996. Anticipation under §102 requires a single reference disclosing every element of the claim as arranged — so the references below are graded on how much of claims 1, 6, 16, 17, 18 they could reach.

Claim-element keys to test against:

• Claim 1: (i) association of address sets with domain identifiers; (ii) derivation of a source domain ID from source address; (iii) a destination domain ID from destination address; (iv) a filtering matrix of entries each holding a forwarding flag; (v) retrieval of an entry as a function of both domain IDs; (vi) forwarding when the flag equals a first value.
• Claim 6: all of the above plus protocol-type equality and source-port-in-range and destination-port-in-range, yielding a protocol index.
• Claims 16 / 17: the formation machinery (address table + 2-D matrix with row de-duplication; protocol table + protocol vector + 3-D matrix with plane de-duplication).
• Claims 18 / 19: method counterpart to claim 1; claim 19 adds that the domain identifier indicates a network protocol.

Blunt assessment on anticipation: On the claim set as I understand it, no reference in my retrieved evidence discloses the full combination of (a) address-set domain identifiers for both source and destination, (b) a protocol/port-derived protocol index, and (c) a filtering matrix indexed by those identifiers with a forwarding flag per entry. That combination — especially the matrix indexing and the claims 16/17 compression steps — is what distinguishes 6,147,976. The realistic §102 exposure is limited to claim 1 / claim 18 (the address-only, broadest claims), and then only if a pre-1996 two-key source/destination prefix-classification lookup with a stored forward/drop action can be produced. Claims 6, 16, and 17 are far more likely to be attacked under §103 than §102.

5. What would close the gap (recommended next queries)

To turn §4 from analysis into evidence, retrieve the actual (56) block:

1. https://patents.google.com/patent/US6147976A/en → the "Patent Citations" table (the backward list).
2. USPTO Patent Center / Public PAIR for application 08/670,733 → "References Cited by Examiner" and the PTO-892 notices (authoritative; shows examiner-cited art and any 892 date).
3. Espacenet "Also published as / Cited documents" for US6147976 (family reported as NONE).
4. USPTO Global Dossier for the file wrapper — any IDS submissions and the examiner's cited-art notices.

Explicit uncertainty statement: I am not certain what references appear on the face of US 6,147,976, and I have not seen its PTO-892. Because I was cut off before completing the backward-citation retrieval, §4 is an independently-constructed prior-art worklist, not the patent's citation record. If you need the certified list, pull it from Patent Center (item 2) rather than from any secondary site.

I'll ground this obviousness analysis in actual prior art. Let me run targeted searches.

Obviousness Analysis — US 6,147,976 under 35 U.S.C. § 103

Scope note on inputs. The task instruction directs me to "use the results from the Prior Art section of this page." No prior-art section was supplied in the materials given to me — the two earlier deliverables (patent summary; litigation/PTAB sections) contain no reference list, and Google Patents' front-page "References Cited" for US 6,147,976 was not retrieved. I therefore built the reference set independently below, anchored on (a) the patent's own admitted prior art in its Background section, which is binding as an admission, and (b) references I retrieved and verified. This is flagged as a gap, not papered over. I also hit a tool-call ceiling before I could retrieve the '976 file history or the examiner-cited art — those should be pulled before any invalidity contention is served.

I build on, and do not repeat, the earlier sections. Two cross-references: the earlier sections establish that the patent expired 2016‑06‑24 and that all five independent claims are untested at the PTAB. Those facts don't change the § 103 analysis, but they do change its use — an obviousness opinion here is a defensive/exposure-sizing tool, not an IPR ground.

1. Framework

Critical framing point. The § 112 ¶ 6 construction is a two-edged sword for the patent owner. It narrows claim scope, but it also means the prior art need only disclose structure performing the recited function — not an identical table layout. Much of the '976's apparent novelty is in arrangement, and arrangement is what § 103 doctrine (and § 112 ¶ 6 equivalents) erodes most readily.

2. The reference set

2.1 Admitted prior art (the patent's own Background)

The specification concedes the following, which I take as admissions against interest:

• Filtering rules "are expressed in existing systems as a table of conditions and associated actions which are applied to each received packet in a predetermined order until a decision to route or drop the packet is reached."
• Existing systems filter on "packet type (TCP, UDP, etc.), source IP address, destination IP address, as well as source and destination TCP/UDP ports," and on ingress/egress router interface.
• Route removal from gateway routing tables was a known alternative (the patent's own characterization of Mogul's "Related Work" passage, below).
• The stated defect is only that "such exhaustive searching of the rule table is time consuming and reduces gateway performance."

That last bullet defines the entire problem the claims solve. The invention is a data-structure substitution for a sequential scan. That framing is dispositive for § 103.

2.2 Prior-art references (verified)

3. Grounds of rejection

Ground I — Claim 1 (and claim 18) obvious over PA‑4/PA‑5 in view of PA‑1, PA‑2, PA‑3

Claim 1 requires, functionally: (a) associate addresses ↔ domain identifiers; (b) retrieve a source domain ID from the packet's source address; (c) retrieve a destination domain ID from the packet's destination address; (d) a filtering matrix of entries each containing a forwarding flag; (e) retrieve one entry as a function of the two domain IDs; (f) forward if the flag equals a first value.

Motivation to combine (articulated reasoning, KSR-compliant):

1. Same field, same problem. PA‑1/PA‑2/PA‑3 and the '976 are all concerned with the identical device — a gateway/router screening datagrams — and the identical defect: sequential rule evaluation at line rate. No field-crossing is required.
2. Known technique to improve a similar device. PA‑4/PA‑5 explicitly present masked, longest-prefix, parallel table lookup as a technique for making a network-layer forwarding decision fast. A POSITA reading PA‑4/PA‑5 with PA‑1 would see that the only difference between a routing decision and a filtering decision is what is stored in the matched entry (an output port vs. a forward/drop bit). Substituting one payload for the other is the epitome of a predictable variation.
3. The patent's own characterization of the problem. The specification states the goal as avoiding the "exhaustive search of all individual rules." Direct indexing by pre-classified field values is the known way to convert an O(rules) scan into an O(1) (or O(#domains)) index. That is not an inventive insight; it is the definition of indexing.
4. Reasonable expectation of success. Both PA‑4 and PA‑6 had already been reduced to practice in gateway-adjacent hardware/software. Nothing in the '976 suggests an unexpected failure mode of the referenced structure.

Predicted patent-owner rebuttal, and response. Owner will argue that PA‑4/PA‑5 are routing references and teach nothing about security policy, and that PA‑1 teaches the very sequential scan the invention replaces. Answer: (i) the '976's own Background admits routing-table route-removal and rule-table filtering as the co-existing prior approaches, making their combination the background of the invention rather than an after-the-fact mosaic; (ii) PA‑1 is not offered to supply the lookup structure, only the policy semantics and the source+destination key — the lookup structure comes from PA‑4/PA‑6; (iii) "a reference must be considered for everything it teaches by way of technology" — PA‑5's "associated information" is agnostic as to what the information means.

Ground II — Claim 1 obvious over PA‑6 in view of PA‑1/PA‑2 (alternative)

PA‑6 is, in substance, a packet filter that precomputes a decision structure offline and field-hashes the packet into it. That is the '976's architecture in functional terms. PA‑1/PA‑2 supply the filtering criteria (source, destination, protocol, ports) and the accept/reject/log semantics. Combining an offline-compiled packet-filter decision structure (PA‑6) with a source-and-destination-keyed screening policy (PA‑1) yields claim 1's function. PA‑7 supplies the further known step of generating that structure from a high-level rule script — directly meeting claims 16/17's "responsive to a command script" limitation.

Why this ground is useful: it removes any argument that the lookup structure must come from the routing arts. Two independent lines of prior art (PA‑4/PA‑5 routing; PA‑6/PA‑7 filtering) each supply the precomputed-structure element.

Ground III — Claim 6 obvious over Ground I/II in view of PA‑1, PA‑2, PA‑3, PA‑7

Claim 6 adds: (a) protocol + source port range + destination port range → protocol index; (b) means for determining that the packet's protocol equals the protocol and its source port is in the source range and its destination port is in the destination range; (c) index into the matrix responsive to the two domain IDs and the protocol index.

Motivation. A network administrator specifying "permit TCP from host A to host B on port 25" will inevitably specify the protocol and port; PA‑7 shows the industry had already generalized this to a "service" object. Once "service" is a first-class criterion and the rest of the mechanism is an indexed table (Ground I/II), including service as an index dimension requires no more than recognizing that a lookup key may have three components. The '976 claims nothing about how the port ranges are matched beyond the numeric comparison inherent in the range definition.

Anticipated rebuttal. The owner may argue that the simultaneity of source-port and destination-port classification (specifically the Protocol Table's dual role and the Protocol Vector's "impossible combination = 0" encoding) is inventive. Response: the impossibility of source-protocol ≠ destination-protocol is a physical consequence of the IP header having one Protocol field — the specification says so itself ("This follows because there is only one protocol field in each packet"). Encoding an impossible combination as zero in a cross-product table is not an inventive act; it is a valid-entry map.

Ground IV — Claims 16 and 17 obvious over Ground I/II/III in view of known table-minimization art

Claims 16 and 17 are the data-structure-building claims, and they are the most vulnerable of the five because their characterizing features are pure table-optimization.

Motivation. The claim 16/17 optimizations exist solely to shrink memory. Shrinking a table by merging duplicate rows/planes and redirecting indices is the canonical table-compression technique, universally known. Moreover, the '976's own specification presents the dedup as an optimization performed on the matrix ("The present invention removes this redundancy"), i.e., a post-hoc engineering step, and provides no unexpected performance or correctness property. Under KSR, "the combination of familiar elements according to known methods is likely to be obvious when it does no more than yield predictable results" — here, a smaller table of identical behavior, which is by definition predictable.

The one feature worth watching. The reflection optimization — detecting Matrix[s][d][p1] == Matrix[d][s][p2] and encoding p2 as −p1, with a lookup that swaps [s]/[d] on a negative value — is a genuinely clever compression. But it is not claimed (the earlier summary already flagged reflection optimization as disclosed-but-unclaimed). It appears only in the specification at step 106. Its presence in the disclosure does not rescue claims 16 or 17, which recite plain row- and plane-deduplication.

Ground V — Claim 18 (method) and Claim 19 (dependent)

Claim 18 is the method counterpart of claim 1 (extract source/destination addresses; identify associated domain IDs; retrieve a matrix entry as a function of both; forward on a first flag value). Grounds I and II apply verbatim — the method steps are the operational description of PA‑4/PA‑5's lookup plus PA‑1's accept/reject action. Claims that "the machine performs X" are no more patentable than the apparatus; offering a method claim alongside a system claim on the same functionality adds no patentable weight (cf. In re Abele/In re Maucorps line; the two are co-extensive in scope here).

Claim 19 (domain identifier indicates a network protocol associated with the address) is met by PA‑7/PA‑2's protocol-and-service classification, applied to the domain-identifier field of PA‑4/PA‑5's table.

4. The five independent claims at a glance

5. Motivation-to-combine summary (for a § 103 articulation)

Four independent, mutually reinforcing rationales supply the required "articulated reasoning":

1. Identical field, identical problem, identical device. PA‑1/PA‑2/PA‑3 and the '976 all address high-throughput gateway screening. No analogous-art or field-of-endeavor dispute is available to the owner.
2. Known technique, known improvement. PA‑4/PA‑5/PA‑6 show that masked, longest-prefix, precomputed table lookup was the recognized method for making routing/filtering decisions fast. Applying a known technique to improve a similar device in the same way is KSR's central holding.
3. Predictable result, no unexpected property. The only benefit the '976 asserts is that "lookup time is dependent only on the number of filtering domains, and not the complexity or number of filtering rules." That is the definition of replacing a scan with an index; it follows necessarily, not unexpectedly.
4. The patent's own framing is a roadmap. The Background states the problem and names the prior approaches; the Detailed Description presents nearly every architectural choice ("an example embodiment," "in an alternative embodiment," "in a most basic form") as a design option. Where the specification itself characterizes alternatives as design choices, the claims cannot claim the choice as invention.

6. Where the obviousness case is weakest

Stated candidly, because an analyst who only builds the reject case is not useful:

• No claim construction exists. "Means for associating," "filtering matrix," and "responsive to" will each require a § 112 ¶ 6 structure identification before any prior-art mapping is final. If the owner successfully argues that the corresponding structure for claims 16/17 is the specific multi-stage pipeline of FIG. 6 (subdomain lists 88/90 → inverse generation → intermediate vector → compression), and that the disclosed algorithm is not merely "merge duplicate rows," the dedup claims become harder. The specification's inverse generation method (scanning rules in reverse order and setting all subdomain entries) is not claimed, so this argument is available only indirectly.
• I have not retrieved the file history. Any examiner-cited art, any § 132 declaration, and any amendment-driven scope surrender are unknown to this analysis. A terminal disclaimer, a narrowing amendment, or prosecution history estoppel could materially change the picture.
• PA‑8 (US 5,835,726) date sensitivity. Its own effective date depends entirely on what the '976-alleged subject matter is supported by the 1993 parent. Use PA‑7 (US 5,606,668, clean 1993‑12‑15 filing) as the primary; treat PA‑8 as supplementary.
• PA‑11 is cited from my own knowledge, not a retrieved document. Verify the Lampson/Saltzer access-matrix citations independently before relying on them.
• PA‑10's "Ranun" name appears in the US 5,606,668 References Cited as "Ranun M.J." — I report it literally as retrieved; the well-known author is Marcus J. Ranum. Do not silently correct the citation in any filing without checking the printed patent.

7. Practical conclusion

On the record I assembled, all nineteen claims of US 6,147,976 are vulnerable under § 103, with claims 16 and 17 the most exposed (pure table-compression) and claims 1 and 18 a close second (a routing-table lookup returning a policy bit instead of a next hop). Claims 6 and its dependents add only subject matter the patent itself concedes was known — protocol, source port, and destination port filtering — arranged as a three-field index.

Three caveats bear on how this should be used:

1. The patent expired 2016‑06‑24 and no assertion is known (per the earlier litigation and PTAB sections). This analysis is therefore an exposure-sizing and demand-letter-response tool, not an IPR ground — filing an IPR on an expired patent buys nothing.
2. The gaps in Section 6 are real. Before any obviousness contention is served, retrieve (a) the '976 front-page References Cited, (b) the file history and any § 132 evidence, and (c) a pre‑June‑1996 source for the pattern/resolution-table indirection teaching.
3. Metadata discrepancy flagged, not corrected: the earlier sections of this analysis are timestamped 2026‑09‑22; this task states 2026‑04‑26. I report both literally.

If the "Prior Art section of this page" referenced in the task instruction exists and contains references I did not have, my ground set should be re-run against it — a reference the examiner already applied would be the single most efficient starting point for any § 103 analysis, and its absence from my inputs is the largest gap in this deliverable.