US 5787071

I'll search for authoritative information on this patent, including any litigation activity.

Let me do a couple more targeted searches to check for any litigation and verify family members.

US Patent 5,787,071 — Analyst Summary

Identification

Note on minor data discrepancies: The full patent text (Google Patents, everypatent.com) gives priority 1994-11-08, filing 1995-11-06, grant 1998-07-28. Unified Patents' portal lists priority 1994-11-07, application 1995-11-05, grant 1998-07-27, expiration 2015-11-05. These are consistent with a one-day/timezone/recordation offset; I have not corrected either source (per the strict-identifier rule), but the dates above reflect the primary patent document.

---

Abstract

"A communication system comprises a plurality of nodes interconnected by links comprising a plurality of connections. The traffic between the nodes is set up by a reserved bandwidth service and/or a non reserved bandwidth service. The non reserved bandwidth service is controlled by a hop by hop backpressure mechanism. When the traffic entering a node exceeds a high threshold, the backpressure mechanism generates stop backpressure primitives in order to throttle the entering traffic. In case of congestion the mechanism is either able to selectively interuppt [sic] the connection contributing to the congestion without affecting the rest of the link traffic, or to globally stop all link traffic. Traffic can be resumed if traffic rates fall below the low threshold values."

---

Plain-Language Overview of the Independent Claims

This patent has two independent claims: claim 1 (an apparatus/system claim) and claim 6 (a method claim). Claims 2–5 depend from claim 1; claims 7–9 depend from claim 6.

Claim 1 — Data traffic flow control system (apparatus)

A congested destination node in a network (nodes linked together, each link carrying one or more logical connections that deliver data cells from a source node into a connection-specific input buffer at the destination) contains two backpressure mechanisms working in parallel:

• First (connection-specific/selective) mechanism — counts how many data cells are sitting in each individual connection's buffer. If a particular connection's buffer exceeds a preset per-connection high threshold, the node issues a connection-specific backpressure signal (a "selective stop") aimed at just that connection.
• Second (global) mechanism — counts the total cells buffered across all connections arriving on a particular link. If that aggregate exceeds a preset global high threshold, the node issues a global backpressure signal for the whole link.

Plain English: Instead of only being able to throttle one culprit connection at a time (which forces you to send a separate message per connection and risks huge in-flight overflow), the node can either single out the misbehaving connection or shut off the entire incoming link with one message when things get bad overall.

Claim 6 — Flow control method

A method performed at a destination node, comprising three steps:

1. Monitor the number of cells stored at any given time in each connection-specific buffer at the destination node.
2. Generate a connection-specific backpressure signal (for forwarding back to the source) for each connection whose stored cell count exceeds a preset connection-specific threshold.
3. Generate a global backpressure signal (for forwarding to the source) if the total stored cells across all connections from that source exceeds a preset global threshold.

Plain English: The method claim recites the same dual-threshold monitoring/signaling scheme as claim 1, but expressed as method steps rather than as hardware mechanisms.

Dependent claim highlights (for context): claim 4 adds low-threshold restart signals (selective restart and global restart); claim 5 recites that backpressure signals are given priority over data cells; claim 8 recites that a selective restart is forwarded only if no global backpressure is outstanding.

---

Litigation / CAFC 2026 Docket Check

No CAFC 2026 docket, district-court case, or PTAB proceeding naming US 5,787,071 was found in the searches performed. Points of caution and uncertainty:

• The patent expired around November 2015, roughly a decade before the 2026 docket window, so active assertion in 2026 is unlikely on its face (though expired patents can still appear in past-damages cases).
• Do not confuse this patent with the "'577 patent" (U.S. 6,377,577) litigated in Cisco Systems v. Arista Networks / ITC Inv. No. 337-TA-945 and in the related Arista 10-K disclosures. That is a different patent number; the Arista filings also reference a '577 patent that "expired on June 30, 2018." Those references are not to US 5,787,071.
• I cannot confirm with high confidence that no such 2026 docket exists; my searches covered Google Patents, Unified Patents, govinfo, courtlistener, and general web queries, but I did not obtain an authoritative USPTO PatentCenter or PACER/CAFC docket listing for this specific number. Treat the "no litigation found" finding as absence of evidence, not proof of absence.

Sources: https://patents.google.com/patent/US5787071/en ; https://www.everypatent.com/comp/pat5787071.html ; https://patentimages.storage.googleapis.com/pdfs/US5787071.pdf ; https://portal.unifiedpatents.com/patents/patent/US-5787071-A ; https://uspto.report/patent/grant/5,787,071

I'll search for litigation involving US patent 5787071.

Let me dig deeper on specific case leads.

Let me check the Unified Patents litigation portal and other litigation databases directly.

Findings on U.S. Patent No. 5,787,071

Patent identified (literally, as requested): US 5,787,071 A — "Hop-by-hop flow control in an ATM network"; application 08/554,113; filed 1995‑11‑06; granted 1998‑07‑28; priority 1994‑11‑08; original assignee International Business Machines Corp.; later assigned to Cisco Systems, Inc. (effective 2000‑01‑26) and then to Cisco Technology, Inc. (effective 2000‑04‑05); expired 2015‑11‑06 (anticipated expiration).

Litigation result: none found

I did not find any litigation — pending or concluded — in which US 5,787,071 was asserted or challenged. Based on the searches I was able to run (Google Patents full‑text/litigation metadata, Unified Patents patent portal, general docket aggregators, RPX/Law360 news), the patent does not appear in any complaint, counterclaim, ITC action, or PTAB/post‑grant proceeding that I could identify.

Important negative/verification notes (so this isn't overread)

1. The Unified Patents portal does have a page for this patent — https://portal.unifiedpatents.com/patents/patent/US-5787071-A — which shows the bibliographic data and current assignee (Cisco Technology Inc., parent Cisco), but the retrieved content did not include any litigation entries. I could not retrieve the page's litigation tab (https://portal.unifiedpatents.com/litigation/caselist) filtered to this patent, and my searches on that portal returned unrelated results. This is a gap, not affirmative proof of absence.

2. A near‑miss that must not be conflated: Diorite Technology, LLC v. Cisco Systems, Inc., Case No. 2:25‑cv‑00591 (E.D. Tex., Judge Robert W. Schroeder III; filed 2025‑05‑28; still Open, first amended complaint filed 2025‑09‑04; claim construction briefing in 2026). That case accuses Cisco products (Broadworks, Webex/Webex Calling, Cisco Desk devices, IP phones) but asserts three former Mitel Networks patents (the ’759, ’421, and ’820 patents) — automatic call‑back, video‑conferencing GUI, and mass audio notification. US 5,787,071 is not among them, and Diorite's asserted patents are unrelated in subject matter and provenance. I flag this only because a keyword search for "5787071" plus "Cisco" can surface it.

3. Other hits for the "5787071" string in my results (a Brazilian JUCE Ceará corporate‑registration PDF, and the ’578 patent references in unrelated bamboo‑scrimber and gaming‑machine opinions) are not this patent and are coincidental numeric matches.

Why an absence of litigation is plausible here

US 5,787,071 is a narrow ATM hop‑by‑hop backpressure patent that was expired as of 2015‑11‑06 — a decade before the current date. It came to Cisco not as a standalone asserted asset but as part of the January 2000 IBM→Cisco Systems transfer (recorded reassignment "ASSIGNMENT OF INTEREST… ASSIGNOR: INTERNATIONAL BUSINESS MACHINES CORPORATION"), with a further intra‑Cisco assignment to Cisco Technology, Inc. Cisco has not been a notable plaintiff asserting legacy ATM flow‑control patents in district court. Note that expired patents cannot be asserted prospectively, though damages for pre‑expiry infringement remain recoverable within the § 286 six‑year look‑back — none of which I found here.

Recommended next steps for a definitive check (I could not complete these)

• PACER / CourtListener RECAP (courtlistener.com) full‑text docket search for "5,787,071" and "5787071" across all districts and the Federal Circuit.
• Unified Patents litigation caselist filtered to US-5787071-A, and the same for portal.unifiedpatents.com/ptab/caselist.
• RPX Insight (insight.rpxcorp.com) patent‑level litigation lookup by the literal number 5787071.
• Westlaw/Lexis LitAlert & KeyCite and Lex Machina patent‑case search — these are the tools that would conclusively confirm a no‑litigation result.

Bottom line: As of this search, no litigation involving US patent 5,787,071 is known — but given that I could not execute every database check above, I'd characterize this as "none found" rather than a certified "none exists." I will not fabricate a case (plaintiff, defendant, jurisdiction, case number, filing date, or status) for a patent where no docket has been located.

I'll verify the structured "no AIA proceedings" signal against public sources before reporting.

Proceedings overview

There are zero AIA trial proceedings on file against US 5,787,071 — no IPR, no PGR, no CBM — and my independent web searches surfaced no petitions, institution decisions, final written decisions, or Federal Circuit appeals involving this patent. The canonical structured source (USPTO ODP, per the "PTAB proceedings on file" block) states no AIA trial proceedings as of the most recent ingest, and targeted searches against PTAB decision repositories, Docket Alarm PTAB dockets, and the Unified Patents patent portal returned nothing for this patent number. The bottom-line defensive posture is unusual but decisive: this is not a "hardened by IPRs" patent, it's an expired, never-IPR'd patent. The '071 patent's term ran 20 years from its 1995-11-06 filing date and therefore ended 2015-11-06 (Google Patents "Anticipated expiration"; the Unified Patents portal lists an expiration date of 2015-11-05, a one-day artifact of a slightly different priority/application-date normalization). Because no AIA trial was ever filed, there is no § 315(e)(2) estoppel, no claim-cancellation record, and no FWD to cite. The defensive value is not "the claims are dead at the PTAB" — it's "the patent is dead by operation of law, and any pre-suit demand issued today is more than a decade past the last possible date of direct infringement."

No proceedings to itemize

I cannot list proceeding entries because fabricating docket numbers is off-limits and none exist. What I can do is document the negative finding and the surrounding enforcement-relevant record, which is what a defendant actually needs:

• Canonical source: USPTO Open Data Portal — no AIA trial proceedings indexed for US 5,787,071.
• Independent verification (2026-09-28): searches of PTAB materials, Docket Alarm PTAB dockets, and the Unified Patents patent page (portal.unifiedpatents.com/patents/patent/US-5787071-A) returned no IPR/PGR/CBM docket, no reexamination, and no litigation tied to this number. I found no Federal Circuit appeal referencing it.
• Caveat: absence of indexed proceedings is strong but not the same as a certified full-docket search. The PTAB's own E2E system (ptacts.uspto.gov) and CourtListener (courtlistener.com) should be run by counsel of record before anyone represents "no proceedings" to a court or an adversary.

Ownership, family, and status (context that explains the silence)

• Original assignee: International Business Machines Corp. Inventors Claude Basso, Jean Calvignac, Daniel Orsatti, Fabrice Verplanken.
• Current assignee: Cisco Technology, Inc. Chain of title: IBM → Cisco Systems, Inc. (recorded 2000-06-14, effective 2000-01-26) → Cisco Technology, Inc. (recorded 2000-04-27, effective 2000-04-05).
• Foreign family: EP 0712220 A1 (status: not_active — Withdrawn); JP 2986085 B2 / JP H08214004 A (not_active — Expired).
• Legal status: Expired – Lifetime (anticipated expiration event logged 2015-11-06); maintenance fees paid at 4, 8, and 12 years.
• Why the PTAB never saw it: the AIA trial regime began on 2012-09-16 (IPR/CBM; PGR for post-AIA patents only). This patent is a 1994-priority, pre-AIA patent that expired in 2015. That leaves only a ~3-year overlap window in which any party had both a statutory vehicle and a live damages incentive — and Cisco, the assignee, was the natural net enforcer of an ATM flow-control patent during a period when ATM was already being displaced by Ethernet/IP. The absence of IPRs here is not a "survived two IPRs" signal; it reflects a patent that was never a commercially potent assertion weapon.

Strategic summary

Claim status: all nine claims are UNTESTED at the PTAB and, as a practical matter, UNENFORCEABLE going forward. Claims 1–5 are apparatus/system claims (claim 1 reciting the dual selective + global backpressure mechanisms; claim 2 the forwarding means; claim 3 the extra-buffer-capacity-above-the-high-global-threshold limitation; claim 4 the restart mechanisms; claim 5 the backpressure-priority-over-data-cells limitation). Claims 6–9 are the method counterparts (claim 6 the monitoring/generating steps; claim 7 the forwarding steps; claim 8 the connection-specific restart conditioned on "no existing global backpressure signal"; claim 9 the global restart). None of these was canceled, narrowed, or even construed by the Board — so there is no FWD reasoning to quote, no surviving-claim list, and no claim-level disposition to rely on. Do not let anyone tell you otherwise.

Estoppel landscape: there is none, and that cuts both ways. Because no IPR/PGR/CBM was ever instituted, § 315(e)(2) estoppel never attached to anyone. In theory that means a defendant could still seek IPR on any ground — but the statute's availability is cold comfort, because the enforcement predicate is gone. The patent expired 2015-11-06, so there is no ongoing direct infringement to enjoin and no future royalty base. Under 35 U.S.C. § 286's six-year damages lookback, even the tail of past damages (which would have run to the expiration date) closed in November 2021. Any demand letter or complaint filed today is attacking a target whose clock ran out twice over. Prior-art grounds remain "available" in the abstract but are largely academic unless someone is trying to invalidate the patent defensively in a declaratory-judgment posture or clear a portfolio cloud.

Pattern signals: no petitioner to find, and no aggressive patent owner. There is no serial petitioner, no defensive aggregator (Unified Patents is not in the chain — its portal merely indexes the patent as Cisco-owned), and no PTAB appeal activity because there was no PTAB activity. The one pattern worth naming is commercial neglect: an IBM-originated ATM flow-control patent, assigned to Cisco in 2000, allowed to expire in 2015 with no AIA challenge and no public assertion campaign surfacing in searches. That is consistent with a patent that was defensive/cross-licensing ballast rather than an assertion asset.

Recommended next steps

• If you are a defendant receiving a demand or complaint citing US 5,787,071: the first response is not an IPR petition — it is 35 U.S.C. § 286 and the expiration date. The patent's 20-year term ended 2015-11-06, and six years later (2021-11-06) the last available damages recovery lapsed. Ask opposing counsel to identify the specific acts of infringement occurring on or before 2015-11-06 and within the statutory lookback; if they cannot, the claim fails as a matter of law.
• Because there is no Final Written Decision, there is nothing to link to or quote. I will not manufacture one. Verify the negative yourself in PTAB E2E / PTACTS and CourtListener before briefing the court, and preserve the search results.
• If your adversary claims an IPR history: demand the proceeding number and the FWD. A real FWD will have a case number of the form IPR20XX-XXXXX and a document on PTAB E2E. If they cite "claims 1–5 canceled" or similar without such a document, that is a factual misrepresentation — the record shows all nine claims of the '071 patent were never subject to an AIA trial.
• If you need invalidity as a hedge (e.g., in a DJ action or to defeat a stale-looking demand): the patent's own cited art is a starting point — US 5,453,981 (Toshiba), EP 0 573 739 A2 (Roke Manor), US 5,455,826 (Özveren), US 5,633,861 (Alcatel Data Networks) — plus the non-patent literature the examiner considered: Gerla & Kleinrock, "Congestion Control in Interconnected LANs," IEEE Network (Jan. 1988), and H.J. Chao, "A General Architecture for Link-Layer Congestion Control in ATM Networks," ISS 1992. These are examiner-cited references, so any IPR ground built on them faces the same art the examiner already weighed — expect to need new art or new combinations.
• Bottom line for the client: there is no PTAB activity to game, and that is the strongest defensive fact in the file. The patent is expired, unasserted in any indexed forum, and owned by Cisco — not a standalone assertable asset. Treat any assertion of the '071 patent as either stale, mis-cited (wrong patent number), or a bluff, and price your response accordingly.

Confidence notes: the zero-proceeding finding rests on the structured ODP block plus my searches and is high-confidence for the indexed record, but should be independently re-run on PTAB E2E before any legal reliance. The 2015-11-06 expiration is derived from the 1995-11-06 filing date in the authoritative patent text; the Unified Patents portal shows 2015-11-05 and a 1994-11-07 priority date, a one-day normalization difference that does not change any conclusion.

I'll research the assignment chain for US 5,787,071 from authoritative sources.

Let me search for the specific assignment records, reel/frame numbers, and correspondents.

Assignment Chain Analysis — US 5,787,071

Important scope note up front: I was able to retrieve the chain (IBM → Cisco Systems → Cisco Technology) and two of the three reel/frame numbers from the Google Patents legal-events record and the patent front matter, but I was not able to pull the live USPTO Assignment Center / Assignment Search abstract of title for this specific patent during this session (the Assignment Center is a JavaScript application that my text-based fetches did not resolve). I therefore have no correspondent-of-record data and no reel/frame for the first assignment. Where a field is unknown I say so rather than guess. Verify against: https://assignment.uspto.gov/patent/index.html (search patent number 5787071) and https://assignmentcenter.uspto.gov/.

---

Inventors

Four named inventors, all residing at IBM France facilities at the time of filing (addresses of record on the patent):

Basis for "employer = IBM": all four assigned their rights to IBM CORPORATION by an assignment recorded 1995-12-29, ~7 weeks after the 1995-11-06 US filing; the La Gaude and Nice sites are IBM's French networking/peripheral development locations. No unusual pattern is evidenced — I found nothing indicating any inventor departed within 12 months of filing, and there is no "portfolio fire-sale" precursor here. Caveat: I could not verify individual departure dates from a primary HR/employment source, so treat the "no departures" observation as absence of evidence.

---

Original assignee

• On the issued patent: International Business Machines Corporation (Armonk, NY). Google Patents lists original assignee as "International Business Machines Corp."
• Products embodying the claims: Yes — IBM shipped ATM switching/routing hardware (the Networking Hardware Division) during the 1990s, and this patent arose from that effort. IBM's networking division, however, was a commercial laggard.
• Primary line of business: diversified computing, software, and services (networking hardware was a small, non-core unit).
• Current status of original assignee: Operating. IBM is an ongoing public company; it exited networking hardware in 1999 by selling the division's patent portfolio to Cisco (see below). No bankruptcy, dissolution, or assignment-of-record to a shell entity is involved.

---

Assignment timeline

Recorded assignments, chronologically. Reel/frame for the 1995 IBM assignment was not retrieved. The two 2000 recordings are drawn from the Google Patents legal-events block for US5787071A (which reproduces the USPTO recordation data, including reel/frame and effective dates).

• 1995-12-29 (recorded) / execution date not stated in the fetched record — Reel /__ (not retrieved)

  • Conveyance: Assignment of interest (inventors → employer)
  • Assignor: Claude Basso; Jean Calvignac; Daniel Orsatti; Fabrice Verplanken
  • Assignee: IBM CORPORATION
  • Correspondent: not retrieved from source (Assignment Center correspondent-of-record not available in this session)
  • Context: standard employee invention assignment to the original corporate assignee, ~7 weeks post-filing.

• 2000-01-26 (effective) / recorded 2000-06-14 — Reel 010832/0878

  • Conveyance: Assignment
  • Assignor: INTERNATIONAL BUSINESS MACHINES CORPORATION
  • Assignee: CISCO SYSTEMS, INC. (a California corporation)
  • Correspondent: not retrieved from source
  • Context: portfolio acquisition — part of IBM's Sept-1999 agreement to sell its networking-hardware patents (reported as ~200 patents) to Cisco Systems; a strategic sale of a whole division's IP, not a targeted single-patent transfer.

• 2000-04-05 (effective) / recorded 2000-04-27 — Reel 010756/0122

  • Conveyance: Assignment
  • Assignor: CISCO SYSTEMS, INC. (a California corporation)
  • Assignee: CISCO TECHNOLOGY, INC. (a California corporation)
  • Correspondent: not retrieved from source
  • Context: internal corporate reorganization — Cisco Systems routinely parks acquired patents in its Cisco Technology, Inc. IP-holding subsidiary; this is a parent→subsidiary housekeeping transfer, not a third-party or NPE transfer. (Note the recording dates here precede the IBM→Cisco Systems recording date even though the effective dates are close; this is a normal recordation-lag artifact.)

No assignment, security agreement, license, merger, or release is recorded after 2000. The 2015-11-06 anticipated expiration is a legal-status event, not an assignment. The only "change of name" event in the family is the EP/JP family counterpart history (EP0712220A1 withdrawn; JP2986085B2), which does not affect the US chain of title.

---

Timeline diagram

timeline
    title Ownership of US 5787071
    1994 : Priority application filed in Europe
    1995 : US application filed
         : Inventors assign to IBM
    1998 : Patent US 5787071 issued
    2000 : IBM assigns portfolio to Cisco Systems
         : Cisco Systems assigns to Cisco Technology
    2015 : Patent expires

---

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT. Neither 2000 assignee is a licensing-only vehicle. Cisco Systems, Inc. and Cisco Technology, Inc. are Cisco's operating and IP-holding entities; Cisco Technology is a wholly-owned subsidiary of Cisco Systems, not a single-member Delaware/Texas shell. No "IP / Patents / Licensing / Holdings" suffix appears in the chain.

2. Known asserter in the chain — NOT PRESENT. The chain runs IBM → Cisco Systems → Cisco Technology. None of these matches a public NPE list (Acacia, Marathon, IV, Wi-LAN/Conversant, Vringo, Pendrell, etc.). The Cisco Technology subsidiary appears in Cisco v. Arista litigation, but the asserted patent there was US 6,377,577 — not US 5,787,071 (do not conflate; flagged in the earlier summary too).

3. Repeat correspondent across the chain — UNCLEAR. I could not retrieve the correspondent of record for any of the three recordings, so I cannot confirm or exclude a recurring attorney/firm. This is a data gap, not a negative finding. Recommend re-running the Assignment Center query and capturing the correspondent field for reels 010832/0878 and 010756/0122.

4. Cascading transfers through chained LLCs — NOT PRESENT. Only two post-issuance transfers, both to entities within the same Cisco corporate family, both in 2000. No LLC-to-LLC hop, no common registered-agent address pattern.

5. Pre-litigation transfer — NOT PRESENT. No infringement suit naming this patent was found in the earlier docket search, and the 2000 transfers long pre-date the (unfound) litigation window. Nothing suggests the chain was arranged to enable assertion.

6. Bankruptcy fire-sale — NOT PRESENT. IBM was not in bankruptcy; this was a solvent strategic divestiture (announced September 1999, part of a $2B alliance). Not analogous to Kodak/Nortel/Polaroid.

7. Privateering — NOT PRESENT. The transfer ran to an operating company (Cisco) that separately migrated IBM's networking customers to Cisco equipment — the textbook opposite of privateering (no transfer from an operating company to an NPE that sues on its behalf). No SEC 10-K/8-K or Patent Progress/EFF coverage surfaced suggesting Cisco used a proxy to assert these IBM patents.

8. Defensive aggregator (anti-NPE) — NOT PRESENT. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. Note, however, that the 1999 IBM/Cisco transaction had a substantial strategic/defensive-aggregation character — Cisco absorbed an entire competitor's networking backlog to foreclose third-party use — even though the record holder is a product company, not a defensive fund.

---

Verdict

Operating-company assertion — with an explicit caveat.

The chain terminates at Cisco Technology, Inc. (a Cisco operating/IP-holding entity), per reel 010832/0878 and reel 010756/0122, and every link is a corporate/strategic transfer (IBM employee-assignment in 1995; IBM→Cisco Systems portfolio sale in 2000; Cisco Systems→Cisco Technology internal reorganization in 2000). Zero NPE signals are present — no shell entity, no known asserter, no cascading LLC transfers, no bankruptcy, no privateering, no defensive aggregator. The caveat is that the strict definition of this verdict requires the owner to be suing on the patent; no assertion of US 5,787,071 was found (and it expired ~2015-11-06), so the label reflects operating-company ownership rather than confirmed active litigation. The only unresolved signal is signal 3 (correspondent recurrence), which is a retrieval gap — worth closing because Cisco Technology is a prolific post-2000 litigation entity elsewhere, but on this specific patent the record is clean.

Verify the chain at: https://assignment.uspto.gov/patent/index.html (search 5787071) — and capture the correspondent field for reel 010832/0878 and reel 010756/0122 to close signal 3.

I'll research the cited prior-art references for US 5,787,071 in detail.

Let me get precise bibliographic details and the substance of the non-patent literature.

Prior-Art Analysis — US Patent 5,787,071 ("Hop-by-hop flow control in an ATM network")

1. Patent-identification confirmation

I matched the identifier 5787071 exactly (no near-number substitution). The record I relied on is the full text of US 5,787,071 A / US5787071A, application 08/554,113, priority 1994‑11‑08 (EP 94 480125.7), US filing 1995‑11‑06, granted 1998‑07‑28, original assignee IBM, later assigned to Cisco Systems/Cisco Technology. This is the same document summarized in the previously generated section, and I did not find any second, distinct granted patent numbered 5787071.

Confidence note / caveat: I could not open a live USPTO PatentCenter or PATENT GRANT full-text page in this session. The authoritative text I used is the USPTO-derived full text mirrored at https://patents.google.com/patent/US5787071/en. The bibliographic matches (number, app. no., dates, assignees, the four "Patent Citations," and the four non-patent citations) are internally consistent, so I treat them as reliable, but I flag that I did not independently retrieve the grant from PatentCenter.

2. The four references cited against 5,787,071

The face of the patent (Patent Citations) lists exactly four patent references and four non-patent citations:

Non-patent citations: Gerla & Kleinrock, "Congestion Control in Interconnected LANs," IEEE Network, vol. 2, no. 1, Jan. 1988, pp. 72–76; and H. J. Chao, "A General Architecture for Link-Layer Congestion Control in ATM Networks," International Switching Symposium 1992, vol. 1 of 2, 25 Oct. 1992, Yokohama, pp. 229–233, XP 000337649.

Anticipating-claim framework

The two independent claims are claim 1 (destination-node system with a first connection-specific backpressure mechanism and a second global, per-link backpressure mechanism, each triggered by a high threshold) and claim 6 (the corresponding method). Dependents: 2–5 (claim 1) and 7–9 (claim 6).

A reference anticipates a claim under 35 U.S.C. § 102 only if it discloses every element. My conclusion, stated up front, is that none of the four cited patent references is a clean § 102 anticipation of claim 1 or claim 6; their relevance is largely as § 103/background art for specific elements. I give the per-element reasons below.

---

3. Reference-by-reference analysis

3.1 US 5,453,981 A — Toshiba

• Full citation: U.S. Patent 5,453,981 (Kabushiki Kaisha Toshiba), "Method of controlling communication network incorporating virtual channels exchange nodes and virtual paths exchange nodes," priority 1990‑10‑16, granted 1995‑09‑26.
• Timing/§ 102 status: Its US application predates the '071 priority (1994‑11‑08); the grant (1995‑09‑26) is only ~6 weeks before the '071 US filing. It is best characterized as potential § 102(e)/§ 102(a) art relative to the '071 priority, not § 102(b) art (grant/printing is not more than one year before 1995‑11‑06). I could not verify the exact US filing date in this session — treat that element as unverified.
• Description: Relates to an ATM network with VP-exchange nodes and VC-exchange nodes. A VC-exchange node has an input port 131 that "monitor[s] the cell flow for each VC arriving from the user terminal," an ATM switch, an output port with per-buffer priority control, and a bandwidth-managing unit 134 that performs VC admission control (number of VCs admitted, VC-to-VP allocation). It estimates "end-to-end transfer quality" for each VC and does bandwidth management / VP shaping. It is fundamentally an admission-control and bandwidth-allocation scheme (calls, VPs, QoS), not a hop-by-hop congestion-relief mechanism.
• Which claims could it potentially anticipate under § 102? None of claims 1–9 as a whole. At most it shows the claim 1/claim 6 preamble environment (nodes interconnected by links; multiple logical connections; per-VC monitoring; per-connection buffering). It does not disclose (a) generating a connection-specific backpressure signal when a per-connection buffer exceeds a high threshold, nor (b) generating a global backpressure signal for a link when the aggregate exceeds a global threshold. There is no teaching of backpressure stop/start primitives returned to an upstream source node.
• Confidence: High that it is non-anticipatory; the reference is clearly about admission control/bandwidth management rather than backpressure.

3.2 EP 0 573 739 A2 — Roke Manor Research (FBRS)

• Full citation: EP 0 573 739 A2 (Roke Manor Research Limited), "Improvements to data transmission systems," GB priority 9212447 filed 1992‑06‑11, published 1993‑12‑15 (A3 published 1998‑05‑06).
• Timing/§ 102 status: Published 1993‑12‑15, i.e., more than one year before the '071 US filing (1995‑11‑06) and before the '071 priority — so it is effectively § 102(b) prior art.
• Description: Discloses the Fast Bandwidth Reservation Shaper (FBRS): "essentially a FIFO buffer with three thresholds (T1–T3), a server and an FRP control function which monitors the state of the queue and its associated thresholds, alters the service rate of the server and handles the FRP protocol." When T1 is crossed the FRP controller "attempts to negotiate a small data rate (R1) with the network"; at T2 it negotiates a higher rate R2; when the buffer empties, allocated bandwidth is cleared to a default. Buffer size L and threshold positions "can be designed … by guaranteeing a maximum delay and a maximum data loss probability." The mechanism is in-call bandwidth re-negotiation via in-band request cells using the Fast Reservation Protocol (FRP), not neighbor-to-neighbor backpressure.
• Which claims could it potentially anticipate? None as a whole. It shares the threshold-crossing concept (T1/T2/T3) that superficially resembles the high/low global and selective thresholds, but the response is to increase the server rate / renegotiate reserved bandwidth, not to emit a connection-specific or global backpressure (stop/start) signal to an upstream node. It likewise does not disclose the dual "per-connection + per-link aggregate" threshold structure of claim 1. Relevant as § 103 art for "buffer thresholds," not § 102 art.
• Confidence: High that it is non-anticipatory; the abstract and description are explicit that the control response is rate/bandwidth negotiation.

3.3 US 5,455,826 A — Oezveren (rate-based flow control) — closest on "two-level accounting"

• Full citation: U.S. Patent 5,455,826 (Oezveren, Cueneyt M.), "Method and apparatus for rate based flow control," filed/priority 1994‑06‑28, granted 1995‑10‑03. (Related: US 5,617,409, "Flow control with smooth limit setting for multiple virtual circuits," Digital Equipment Corp.)
• Timing/§ 102 status: US application filed 1994‑06‑28, before the '071 priority; granted 1995‑10‑03 (within one year of the '071 filing). Best characterized as potential § 102(e) art; not § 102(b).
• Description: A flow-control circuit located in the transmitting node. A shared bandwidth pool on the link is shared among the connections, and each connection has a connection-specific bandwidth allocation guaranteeing a minimum rate (Min/T). The circuit operates in periodic "epochs" and contains: a global counter (DTUs sent on all connections during the epoch) compared against an upper threshold UT; and a per-connection counter for each connection. If the global counter exceeds UT, a global limit register is set to Min, and any connection whose counter exceeds the limit is disabled for the remainder of the epoch. So this reference does disclose a per-connection measure plus an aggregate/global measure compared to thresholds, which is conceptually parallel to the '071's two levels of monitoring.
• Which claims could it potentially anticipate? None of claims 1–9 as written. Critically, (i) the mechanism is in the transmitting (source) node, not the destination-node input buffers ("receive buffer" here is a receiving-node storage unit, and the counters live in the transmitter); (ii) the control action is to throttle the sender's own rate for the epoch — there is no "backpressure signal" generated by a destination and forwarded to a source; and (iii) there is no per-link global stop/start primitive. It does not disclose the "connection-specific input buffer at a destination node," the "connection-specific backpressure signal," or the "global backpressure signal for the particular link." At most it is § 103 art that could be combined to suggest per-connection + global threshold accounting.
• Confidence: High that it is non-anticipatory of the independent claims; medium confidence in the finer point of whether its "receive buffer" language could be stretched toward claim 1's "connection-specific input buffer at a destination node" — the counters are still transmitter-side, so the stretch fails.

3.4 US 5,633,861 A — Alcatel Data Networks — likely not § 102 art at all (date problem)

• Full citation: U.S. Patent 5,633,861 (Alcatel Data Networks Inc.), "Traffic management and congestion control for packet-based networks," priority 1994‑12‑19, granted 1997‑05‑27. (EP family member EP 0 719 012 A2, published 1996‑06‑26.)
• Timing/§ 102 status — flag this: Its US filing/priority is 1994‑12‑19, which is after the '071 priority date of 1994‑11‑08 (and after the earliest invention date tied to that EP priority). Its own publication events (US grant 1997‑05‑27; EP publication 1996‑06‑26) are also after the '071 filing. On those dates it is not § 102(a) or § 102(e) prior art relative to the '071, and it cannot be § 102(b) art. Its appearance in the "Patent Citations" list is therefore noteworthy — it may have been submitted in an IDS or cited for general relevance rather than as anticipatory art. (I have not located the examiner's stated basis; treat this as an observation, not a proven fact.)
• Description (for completeness): End-to-end, feedback-based congestion management. Each node measures the utilization of critical resources (CUF); utilization information is piggy-backed on packets flowing in the backward direction from destination to source (with special empty packets generated when needed); the source increases or reduces its packet submission rate based on the maximum resource utilization along the forward path.
• Which claims could it potentially anticipate? None. It is an end-to-end rate-adaptation (EFCI-style) feedback scheme, not hop-by-hop backpressure, and it has no per-connection and per-link dual-threshold stop/start signals. Even if it were § 102 art, it does not disclose claim 1/claim 6 elements.
• Confidence: High on non-anticipation; high on the date-based conclusion that it postdates the '071 priority (the date 1994‑12‑19 vs. 1994‑11‑08 is clear), but I did not verify the examiner's rationale for citing it.

---

4. Non-patent literature

• H. J. Chao, "A General Architecture for Link-Layer Congestion Control in ATM Networks," ISS 1992, vol. 1, pp. 229–233 (25 Oct. 1992). Of the eight references, this one is, on its title and venue, the most conceptually on-point for the "global/link-level" half of claim 1 (link-layer congestion control in ATM). It predates the '071 priority (Oct. 1992 < Nov. 1994), so it is § 102(b)-type material. However, I could not retrieve its full text in this session, so I cannot responsibly assert it discloses the specific "connection-specific buffer high threshold → connection-specific backpressure signal" plus "per-link aggregate → global backpressure signal" combination. Treat it as the highest-priority item to pull and read in full; do not treat the above as a finding that it anticipates.
• Gerla & Kleinrock, "Congestion Control in Interconnected LANs," IEEE Network, Jan. 1988, pp. 72–76. Classic background on backpressure/congestion control in interconnected LANs (and, per the patent's own Background Art, the genre in which simple backpressure is discussed, alongside Rodeheffer's "Experience with Autonet"). It is § 102(b)-type background material but, based on the title/venue and the patent's own characterization, is directed to simple/non-selective backpressure — i.e., it cuts against anticipation of the selective + global dual mechanism rather than supporting it. Full-text not retrieved; confidence limited.

---

5. Overall conclusion

1. No § 102 anticipation of claim 1 or claim 6 (or their dependents 2–9) is established by any of the four cited patent references. Each fails on at least one essential element — most commonly the requirement of a destination-node-generated connection-specific backpressure signal and a separate per-link global backpressure signal.
2. Most relevant cited patent references, in order:

• US 5,455,826 — closest on the "per-connection accounting + global accounting against thresholds" idea, but transmitter-side, rate/epoch-based, with no backpressure signaling (best § 103 candidate).
• EP 0 573 739 A2 — closest on "multi-threshold buffer with high/low crossings," but the control response is bandwidth renegotiation, not backpressure.
• US 5,453,981 — relevant only to the generic ATM per-VC monitoring/admission-control environment.
• US 5,633,861 — least relevant and appears to postdate the '071 priority, so likely not § 102 art at all.

1. The NPL citation most likely to be genuinely material is the Chao ISS 1992 link-layer congestion-control paper, which I could not verify in full here; if researching further, retrieve and read it before making any § 102 assertion.
2. Flagged contradiction/uncertainty to carry forward: (a) US 5,633,861's 1994‑12‑19 filing date is after the '071's 1994‑11‑08 priority, which is inconsistent with its treatment as anticipatory art; I could not confirm the examiner's citing rationale. (b) I could not retrieve the exact US filing date of US 5,453,981 or the full texts of the two NPL items. (c) Live USPTO PatentCenter/PACER retrieval was not obtained; the full text used is the USPTO-derived mirror at patents.google.com. These are absences of verification, not proof of any contrary fact.

(This analysis builds on, and does not repeat, the previously generated identification, abstract, claim-overview, and litigation sections of this patent.)

I'll research the prior-art references cited on this patent's face before building the §103 analysis.

Let me research the remaining prior-art references and the non-patent literature.

Let me look for the two non-patent references and the EP counterpart's search report.

Obviousness Analysis — US 5,787,071 under 35 U.S.C. § 103(a)

Builds on the previously generated Identification, Abstract, and Claim-Overview sections. Prior-art content below is drawn from the "Patent Citations (4)" / "Non-Patent Citations (4)" / "Citations (4)" blocks of the patent page, supplemented by reference-level research where noted.

---

1. Bottom line

Claims 1–9 are, on the record as it stands, vulnerable to a § 103(a) rejection. The patent's own Background section supplies two of the three legs of the rejection by admission, and the two non-patent references cited on the face of the patent collectively supply the third. The most dangerous single reference is H. J. Chao, "A General Architecture for Link-Layer Congestion Control in ATM Networks," ISS 1992, Yokohama, pp. 229–233 (XP000337649) — which the EPO cited as an "[X]" category document (i.e., relevant alone) against the priority-matched EP family member EP0712220A1. The commercial and claim-drafting history also matters: the genuinely novel subject matter in the specification (dynamic per-connection threshold adjustment as a function of NAC, multi-slot control-cell packing, latency/discard counters) is not claimed, which sharply narrows the non-obviousness argument available to the patentee.

That said, claim 5 ("backpressure signals are given priority over data cells") is the weakest link in the rejection case on the cited art alone, and the § 102(e) status of US 5,633,861 carries a real date problem (§ 5 below).

---

2. Governing law and critical dates

Framework applied: Graham v. John Deere Co., 383 U.S. 1 (1966) (scope/content of prior art; differences; PHOSITA level; secondary considerations) and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) ("the combination of familiar elements according to known methods is likely to be obvious when it does no more than yield predictable results"; the reason to combine may be found in design incentives, market forces, or "any need or problem known in the field"). MPEP §§ 2141–2143 (exemplary rationales (A)–(G)) supplies the articulation framework.

A separate, and here decisive, line of authority: statements in an applicant's own specification are admissions usable as prior art under § 103. MPEP § 2129; In re Nomiya. The '071 Background section is unusually generous to the examiner in this respect — see § 5.

---

3. Person of ordinary skill in the art (POSITA)

For a 1994–95 ATM flow-control invention: a B.S. in electrical engineering or computer science (or equivalent) plus 2–4 years of experience in packet/ATM switch design or network protocol development, or an M.S. with 1–2 years; working familiarity with B-ISDN/ATM Forum traffic-management work, credit-based vs. rate-based flow control, per-VC queuing and round-robin scheduling, and buffer threshold/hysteresis design. This is a mature, crowded art: the '071 itself acknowledges that "a backpressure concept … has been extensively studied and is already implemented," and the face of the patent cites two 1988–1992 non-patent references plus four patents. A high skill level in a crowded art cuts against non-obviousness.

---

4. The prior art of record — capsule disclosures and status

Honesty flag. I have not obtained and read the full texts of Chao (XP000337649) or Gerla/Kleinrock. My characterisation of Chao rests on its title, its citation as an [X] reference against the sibling EP application, and its repeated citation in the same technical role in other ATM congestion-control cases (e.g., EP0707399A1, EP0647081B1). The specific-disclosure column for those two references must be verified against the full papers before the rejection is reduced to writing. By contrast, the four patent references' disclosures above are drawn from their own specification text.

---

5. What the claims actually require — and, importantly, what they do not

This is the pivot of the whole analysis. Comparing the claim set against the specification:

Claimed:

• (C1/C6) Per-connection and per-link aggregate monitoring of buffered cells at a destination node, with a connection-specific high threshold and a global high threshold, each producing a corresponding backpressure signal.
• (C2/C7) Forwarding of those signals to the source; source stops the specific connection, or all connections to that node.
• (C3) Buffer headroom above the global high threshold ≥ cells already in flight.
• (C4) Low-threshold restart signals (per-connection and global).
• (C5) Backpressure signals prioritised over data cells.
• (C8) Selective restart forwarded only if no global backpressure is outstanding.
• (C9) Global restart on falling below a global low threshold.

Expressly disclosed but NOT claimed (all in the specification): dynamic adjustment of the selective threshold as a function of the number of active connections NAC (FIG. 14); the 1-to-12 selective signals packed into one private ATM cell (FIG. 15); the 8%/one-in-twelve reserved-bandwidth allocation for control cells; the PBPXCNT/CBPXCNT latency counters and PD/SD discard bits (FIGS. 6, 8–13); the round-robin scheduler removal/reinsertion mechanism (FIG. 7).

Consequence: the patentee's best non-obviousness story lives in the unclaimed disclosure. Under § 103 the inquiry is limited to the claimed subject matter (Graham step 2; KSR), and the Federal Circuit has repeatedly refused to let unclaimed features supply the inventive step. The claimed subject matter is, stripped down, "use two nested thresholds — a per-flow one and an aggregate one — instead of just one, and add hysteresis."

§ 102(e) caveat on US 5,633,861

US 5,633,861's US filing date (1994-12-19) is after the '071's EP priority date (1994-11-08) but before the '071's US filing date (1995-11-06). Because In re Hilmer bars an applicant from using a § 119 foreign priority date to antedate a § 102(e) reference, US 5,633,861 does qualify as § 102(e) art against the '071's 1995-11-06 filing. But if the patentee can swear behind the reference under Rule 131 by establishing an actual date of invention before 1994-12-19 — very plausible given the EP application was already on file on 1994-11-08 — US 5,633,861 drops out entirely. I would not build a rejection on US 5,633,861 alone. It is best used as corroborating art, or as secondary evidence of the state of the art.

---

6. Combination A (primary): Chao + Gerla & Kleinrock

Proposed rejection: Claims 1–4, 6, 7, and 9 obvious over Chao in view of Gerla & Kleinrock (and, as to claims 4 and 8, further in view of Oezveren '826 or EP 0573739 A2).

Mapping:

• Chao supplies a link-layer, hop-by-hop congestion-control architecture for ATM — which, by the very nature of link-layer control between adjacent nodes, means a node monitors what it has accepted from its upstream neighbour(s) and returns control information to that neighbour. Chao's "general architecture" is the element the '071 presents as its inventive contribution. Its [X] citation against EP0712220A1 — the priority-matching EP case sharing the same inventors and this specification — is strong contemporaneous evidence that an examiner of ordinary skill regarded Chao as reaching the claimed subject matter standing alone.
• Gerla & Kleinrock supplies the express teaching that congestion control at a node must be exercised at two granularities — control directed to the source(s) contributing to congestion versus control of the aggregate — which is precisely the selective-vs-global duality of claim 1. (This is the same duality the '071 Background describes when it criticises Rodeheffer's non-selective backpressure for lacking "fairness" and Vickers & Suda's selective approach for requiring "prohibitively large" buffers.)
• Claims 4 and 9 (restart on low thresholds): supplied by EP 0573739 A2 (T1–T3 threshold crossings with stepwise service-rate change and clear-down on empty) — a § 102(b) printed publication directly on point as to buffer-threshold hysteresis in ATM — or by the '071's own admission that backpressure signals "stop or start" traffic (Rodeheffer, admitted).

Motivation to combine (articulated rationales):

1. MPEP 2143(A) / KSR — known elements, known functions, predictable result. Both references are in the same field (ATM/wide-area data congestion control), address the same problem (cell loss at a congested node), and address it in the same way (backward control messages to upstream nodes). Their combination is that of two known flow-control granularities, each performing exactly its known function.
2. MPEP 2143(C) / KSR — known technique applied to a similar device. The art already controlled a link; the art already controlled a connection; applying both to the same node buffer is improvement-by-parallel-mechanism, an "obvious to try" step where the prior art contains a finite number of identified, predictable solutions.
3. MPEP 2143(D) / KSR — known device "ready for improvement." The '071 itself frames the inventive problem as: Rodeheffer's global backpressure is unfair; Vickers & Suda's per-connection backpressure alone blows the buffer budget because of in-flight cells. Where the patentee has articulated the problem this crisply, the solution — do both, and let the aggregate mechanism house-clean when the selective mechanism is too slow — is the predictable design choice. KSR: "any need or problem known in the field … can provide the reason."
4. Design incentive / multicasting of control overhead. One global control message replaces N per-connection messages — an explicit and well-recognised bandwidth-saving incentive in this literature (Oezveren '826) and echoed in the '071's own stated objective of "generat[ing] reduced traffic control."

A POSITA with a reasonable expectation of success (the mechanisms are orthogonal and do not interfere; the only engineering questions — how much headroom, where to set the low threshold — are routine and are in fact the subject of the unclaimed portion of the '071) would implement this combination.

---

7. Combination B (for claims 4 and 8): + Oezveren '826

US 5,455,826 is the single most useful secondary reference because it teaches a hierarchical/nested limit structure: a connection counter per connection, a global counter and global limit register for the aggregate over the link, and — important for claim 8 — operation in which the global limit overrides the per-connection limit once the aggregate threshold is crossed ("if the resulting value of the global counter exceeds UT, the value of the global limit register is set to Min"; "if the connection counter exceeds the value in the global limit register, further transmissions on that connection are disabled"). Claim 8's requirement that a selective restart be forwarded only if no global backpressure is outstanding is the grant-side mirror of that same arbitration: the global condition is senior to the per-connection condition.

The motivation is not "combine two references" so much as recognise the same priority rule on the other side of the interface — a one-line design decision for a POSITA, and the same decision the '071's own FIG. 13 describes ("The upstream node will resume scheduling cells from connection Ck … unless the link comprising connection Ck is subject to a global backpressure condition").

Also note US 5,633,861 teaches piggy-backing control information onto return packets and generating special empty return packets when no return traffic exists — useful corroboration that dedicating return-path capacity to control signalling was routine, which is the general backdrop for claim 5.

---

8. Combination C (alternative, uses only the examiner's own set): Toshiba '981 + Roke Manor '739 + Oezveren '826

If Chao is challenged (e.g., the patentee argues non-analogous art or contests the paper's disclosure), the same result is reachable from the four references the examiner actually cited:

This is the classic KSR "combination of familiar elements yielding predictable results" posture, and it has the practical advantage that the examiner has already found these references and found them relevant to this exact family.

---

9. Combination D: the applicant's-admission route (independent of the NPL)

The '071 Background section places the following on the record as admitted prior art (MPEP § 2129):

1. Rodeheffer, "Experience with Autonet," Computer Networks and ISDN Systems, vol. 25, 1993, pp. 623–629 — admitted as describing a hop-by-hop backpressure mechanism in which "a backpressure signal … stops or starts the traffic coming from the upstream nodes," and in which "every source is supposed to be able to stop or restart its NRB traffic when receiving an appropriate backpressure signal." This is claims 1 (global mechanism), 2, 6 and 7 in substance.
2. B. J. Vickers and T. Suda, "Connectionless Service for Public ATM Networks," IEEE Communications Magazine, Aug. 1994, pp. 34–42 — admitted as describing "a selective congestion control mechanism" and "a scheme through which best effort service may be provided in ATM networks using hop by hop flow control," in which the node "must be selective and must operate only on those connections that are contributing to congestion." This is claim 1's first backpressure mechanism in substance.
3. The applicant's own statement of the problem to be solved: the unselective mechanism "is unable to provide fairness"; the selective mechanism "has to [throttle] one connection at a time," so that because of in-flight cells "the number of ATM cells which have to be received into buffers after the backpressure signal has been sent but before the traffic actually stops, can be very large," and "in order to guarantee no traffic is lost, the required buffer space would become prohibitively large, which is incompatible with an efficient hardware implementation."

That third admission is the motivation to combine in the applicant's own words. Adding the global mechanism to the selective one, so that one message can throttle the whole link when the per-connection messages cannot keep up, is the direct and predictable response to the problem the applicant itself identified. KSR: "the combination of familiar elements according to known methods is likely to be obvious." And the remaining claim elements — headroom ≥ in-flight cells (claim 3) and low-threshold restart (claims 4, 9) — are the very things the applicant says are needed to make a loss-free hop-by-hop scheme work, i.e., they are the routine engineering consequences of the admitted design, not inventive additions.

Note one drafting trap for the patentee: the '071's own description, describing the global threshold constraint, states that "the extra buffer capacity above the threshold is greater than the aggregate latency of the upstream nodes." That is a direct written description of claim 3. Claim 3 therefore risks reading on the admitted-art requirement itself, not on any new discovery.

---

10. Claim-by-claim disposition

Nested-claim observation: because claims 2–5 and 7–9 all depend on 1 or 6, the entire claim set falls with the two independents. There is no claim whose validity survives an adverse finding on claims 1 and 6.

---

11. Anticipation/obviousness note — the EP [X] finding

The EPO's Global Patent Index entry for EP0712220A1 (same priority date 1994-11-08, same inventors, same specification) records:

[X] H.J. CHAO: "A General Architecture for Link-Layer Congestion Control in ATM Networks.", INTERNATIONAL SWITCHING SYMPOSIUM 1992, vol. 1 of 2, 25 October 1992, YOKOHAMA, pp. 229–233, XP000337649

Category [X] means the EPO regarded Chao as relevant alone — i.e., sufficient to defeat novelty or inventive step without combination. EP0712220A1 was thereafter withdrawn (never granted), while the US case issued as the '071. That divergence cuts in two directions and should be flagged rather than assumed away:

• For the challenger: an examiner applying the same reference to the same disclosure reached an [X] finding. That is contemporaneous evidence of what a skilled examiner considered the art to teach.
• For the patentee: the US claims were allowed over a record that included Chao, and it is possible the US claims are narrower than the EP claims as filed, which would explain the divergence. The US claim set's principal narrowing relative to the disclosed invention is that it recites thresholds and signals, not the NAC-based dynamic threshold rule, the control-cell format, or the latency counters — narrowing that, if anything, helps the challenger, since it removes the features that would have distinguished the claims.

I could not locate the EP claim set as filed, so I cannot verify what EP0712220A1 originally claimed, nor whether the [X] citation was made against its independent claims as filed. Treat the "US issued, EP withdrawn" fact as a lead requiring file-history retrieval (USPTO PatentCenter / EPO Global Dossier), not as a conclusion.

Prosecution-history recommendation: pull the '071's US file wrapper. The two NPL references were cited by the examiner, so there should be an examiner's reasons-for-allowance or an interview record articulating why Chao (cited) was not treated as anticipating the US claims. That statement — or the absence of one — will largely determine whether a reexamination or IPR petition is worth filing.

---

12. Secondary considerations (objective indicia)

No objective indicia appear on the record assembled here, and the patentee would face a nexus problem on all of them:

• Commercial success — the patent was assigned IBM → Cisco Systems (2000-06-14) → Cisco Technology (2000-04-27). Cisco's ATM/broadband success is not probative without a demonstrated nexus to the claimed dual-threshold mechanism specifically (and KSR cautions against weak nexus arguments).
• Long-felt need / failure of others — the applicant's own Background describes the Rodeheffer and Vickers/Suda approaches as already implemented and working; the record shows incremental refinement, not a long-felt unmet need. This is a weakness, not a strength, of the patentee's position: the '071 characterises the prior art as "extensively studied and already implemented."
• Licensing — the IBM→Cisco assignment is a corporate transaction; the record shows no royalty or licensing program attributable to the claimed subject matter.
• Expiry — anticipated expiration 2015-11-06; the patent is Expired – Lifetime. This is relevant to remedies (rolled-over past damages only, within § 286's six-year lookback from complaint), not to validity. It also reduces the practical likelihood of a 2026 validity challenge, as noted in the earlier Litigation section, but it does not affect the legal analysis.

A teaching-away argument is theoretically available (one could argue the art taught either per-connection or aggregate control, not both), but it is undermined by the same Background section: the applicant never asserts that the art taught away from combining the two. At most the applicant asserted the two prior approaches were deficient in isolation — which is the classic setup for an obvious combination, not for teaching away.

---

13. Where the rejection is weak, and what would be needed

Disclosed but unclaimed — do not over-read the specification. The '071's specification contains subject matter that is genuinely non-trivial: the NAC-indexed threshold adjustment rule of FIG. 14 (with its worked analysis showing 50% → 97% → 100% buffer utilisation), the 1-to-12 selective-signal packing into a private ATM cell (FIG. 15) with a dedicated VPI/VCI and PTI = 110, the ~8% reserved-bandwidth carriage of control traffic, and the PBPXCNT/CBPXCNT latency counters with PD/SD discard bits. None of that is claimed. If a challenger drafts the rejection around the specification rather than the claims, it will be met with the response that the cited art does not teach these features — and that response is legally irrelevant to § 103 so long as the claims do not recite them. Keep the mapping strictly to claim language.

The genuine soft spot — claim 5. "Backpressure signals are given priority over data cells in data flows between nodes" is not squarely taught by any of the four cited references. US 5,633,861's special empty return packets and OAM-era control-cell practice are suggestive but not direct. If claim 5 is to be attacked, additional art (ATM OAM cell priority, or a switch-egress prioritised-control-queue reference) should be located. Note, however, that claim 5 is dependent; invalidating it is not necessary to clear the set.

The Nachteil of the Hilmer-dependent reference. US 5,633,861's only available basis is § 102(e) as of 1994-12-19, and the patentee can very plausibly antedate it via Rule 131 given the 1994-11-08 EP filing. Do not treat US 5,633,861 as load-bearing.

Verification still required (I flag this explicitly rather than assert it):

• Full text of Chao, XP000337649 — particularly whether it discloses (a) separate per-connection and per-link aggregate thresholds at the same node, and (b) backward stop and start signalling. The [X] citation is documented; the specific disclosure is not, on the record I have.
• Full text of Gerla & Kleinrock at p. 73 — the selective/global duality.
• The '071 US file wrapper (reasons for allowance over Chao); EP0712220A1 as-filed claims and the EP withdrawal record; JP2986085B2 counterpart prosecution (JP examiners frequently cite additional art).

---

14. Summary judgment for the analyst

The '071 is a strong § 103 target as to claims 1–4 and 6–9, on a combination that requires no new art beyond what is already on the face of the patent:

Primary: Chao (ISS 1992, XP000337649) in view of Gerla & Kleinrock (IEEE Network 1988) — with the applicant's own Background admissions (Rodeheffer; Vickers & Suda) supplying the express problem statement and motivation, and Oezveren '826 / EP 0573739 A2 supplying the low-threshold-restart and hierarchical-limit elements.

Fallback: the examiner's own four references (US 5,453,981 + US 5,455,826 + US 5,633,861 + EP 0 573 739 A2) on the KSR "familiar elements, predictable results" rationale.

Claim 5 is the only claim for which I would not presently assert a strong rejection on the cited art alone.

---

Sources

• US 5,787,071 (full text, claims, citations, family): https://patents.google.com/patent/US5787071/en and https://patentimages.storage.googleapis.com/pdfs/US5787071.pdf
• US 5,453,981 (Toshiba): https://patents.google.com/patent/US5453981
• US 5,455,826 (Oezveren, rate-based flow control): https://patents.google.com/patent/US5455826A
• US 5,633,861 (Alcatel Data Networks): https://patents.google.com/patent/US5633861/en ; EP family EP0719012A2
• EP 0 573 739 A2 (Roke Manor Research): https://patents.google.com/patent/EP0573739A2 ; https://data.epo.org/gpi/EP0573739A3-IMPROVEMENTS-TO-DATA-TRANSMISSION-SYSTEMS.html
• EP 0 712 220 A1 (family member; [X] Chao citation): http://data.epo.org/gpi/EP0712220A1
• Chao, ISS 1992, cited in: https://patents.google.com/patent/EP0707399A1/en ; https://patents.google.com/patent/EP0647081B1 ; NYU Scholars record: https://nyuscholars.nyu.edu/en/publications/
• Unified Patents portal (dates variant): https://portal.unifiedpatents.com/patents/patent/US-[5787071](/patent/5787071)-A
• everypatent mirror: https://www.everypatent.com/comp/pat5787071.html

Note: the current task header states the date as April 26, 2026; my operating date is 2026-09-28. Nothing in the analysis above is date-sensitive beyond the 2015 expiration already noted in the Litigation section.