US 7305585

I'll search for authoritative records on US 7,305,585, including any 2026 CAFC activity.

Let me check for any litigation or assignment records, and confirm the bibliographic data.

Let me do a final check for any 2026 activity or reassignment involving the assignee.

Summary — US 7,305,585 B2

Reading note / search-scope caveat: I could not query USPTO PatentCenter, PAIR, or the CAFC docket system directly from this session. The bibliographic data below is taken from the authoritative full patent text supplied to me (Google Patents rendering of the granted patent), cross-checked against Justia and the granted-patent PDF. Anything I could not corroborate is flagged explicitly as uncertain.

Bibliographic data

Term / status discrepancy worth flagging. The Google Patents header states "Expired - Fee Related, expires 2024-11-18," but the prosecution/legal-events record shows: 4-year fee paid 2011-05-11 → maintenance-fee reminder mailed 2015-07-17 → lapse for failure to pay maintenance fees 2015-12-04 → "patent expired due to nonpayment of maintenance fees under 37 CFR 1.362" (effective 2015-12-04). So the ~2024-11-18 figure appears to be the statutory term (20 years from 2003-05-23 plus adjustment), while the patent actually ceased for fee nonpayment in December 2015. Treat the enforcement window as closed since 2015.

Abstract (as issued)

Apparatus and methods to improve the speed, scalability, robustness and dynamism of data transfers to remote computers across a network are provided. The fully distributed data transfer and data replication protocol of the present invention permits transfers that minimize processing requirements on master transfer nodes by spreading work across the network. The result is higher scalability than current centralized protocols, which are static and cannot guarantee retention of replicated data and files. The present invention further provides for more dynamism and allows fault-tolerance by distribution of functionality. The ability to distribute data is simplified through asynchronous and autonomous data transfer.

Independent claims in plain language

• Claim 1 (method). A method for transferring data between networked computing devices with two steps: (a) transferring the data using a broadcast transfer protocol that is sessionless — i.e., no prior registration/connection/handshake or per-node session state is required; and (b) symmetrically recovering corrupted or missing data from any one of the networked computing devices (not necessarily the original sender), where that recovery is fully asynchronous and autonomous relative to the transfer. In essence: bulk multicast push with no session, plus a decoupled peer-to-peer repair phase in which any participating node can supply the missing bytes.

• Claim 20 (apparatus). A computing device configured to perform the same two operations — sessionless broadcast transfer of data, and symmetric recovery of corrupted/missing data from any one of a plurality of networked devices, the recovery being fully asynchronous and autonomous relative to the transfer. (This is a "configured to" device claim, not a means-plus-function claim; the first limitation is the sessionless broadcast transfer, the second is the symmetric/asynchronous recovery.)

• Claim 25 (computer-readable storage medium). A CRM carrying a program executable by a computer to perform the claim 1 method — i.e., sessionless broadcast transfer plus fully asynchronous, autonomous symmetric recovery.

Dependent-claim highlights (context for scope)

Claims 2–19 depend from claim 1: catch-up transfer to a node introduced to the network after the original transfer (2); completing interrupted transfers (3); removing partial files, including at a preset life span, or after abort/cancel from user request or system crash (4–8); delta/subset transfers of only modified portions with a per-node "deltas" log (9–12); multiple concurrent transfers through a single transfer/recovery instance (13); point-to-point exchange processes in recovery (14); throttling network traffic (15); fault-tolerant client interface, user- or tool-launched (16–18); and constant, configurable network traffic via an adaptive distributed selection mechanism (19). Claims 21–24 and 26–28 narrow the device and CRM claims to the adaptive selection mechanism and to node types (PC, PDA, cellular phone).

Related family (for context)

• US 11/926,064 — continuation → US 7,707,457 B2, "Completing an interrupted data replication operation"
• US 10/893,752 — continuation-in-part (published US 2005/0060608 A1)
• US 11/067,458 (published US 2005/0216910 A1) and US 12/045,165 (published US 2008/0222234 A1)
• EP family member: EP 1365538 A1, "Implementing a scalable, dynamic, fault-tolerant, multicast based file transfer and asynchronous file replication protocol"

CAFC / litigation check — no 2026 activity found

I ran targeted searches for Federal Circuit 2026 dockets, IPRs, and district-court litigation involving "7,305,585." I found no Federal Circuit appeal, PTAB proceeding, or 2026 CAFC docket referencing this patent. The record evidence also shows no litigation history for it (unlike its continuation). Because I could not query the CAFC docket or Docket Navigator directly, I state this as no activity located, not as a certified negative.

Explicit anti-conflation warnings — do not mix these up with 7,305,585:

• US 7,075,585 (CF CRESPE LLC) — broadband television receiver patent litigated in MaxLinear, Inc. v. CF CRESPE LLC (Fed. Cir. 2018, IPRs 2015-00728 / 2015-00615 / 2017-00592). A different patent; the shared trailing digits are coincidental.
• US 6,730,585, US 7,290,864, and the "'558 patent" in the Apple v. Qualcomm ITC matter — all unrelated.
• A numeric hit for "7305585" in a pharmaceutical compound table (Trimoprostil / JP registry number) is a coincidental ID match, not this patent.
• A third-party aggregator page lists Benoit Marchand with "15 patents at Stmicroelectronics Sa" including 7305585. Given the assignment-of-record to eXludus Technologies Inc. (Canada), that assignee attribution appears unreliable and should not be relied on.

Uncertainty statements. (1) The exact PTA-calculated statutory expiration vs. the actual 2015 fee lapse should be confirmed in PatentCenter if a precise date matters. (2) Claim 20's full text as rendered in the supplied document is abridged in one snippet; the operative limitations are as I described them, but retrieve the printed claim set if exact wording is needed. (3) I did not independently verify the EP 02011310.6 priority document's contents.

Grounding: https://patents.google.com/patent/US7305585/en; https://patents.justia.com/patent/7305585; https://patentimages.storage.googleapis.com/92/19/5a/13a61e7fdee488/US7305585.pdf

I'll search for litigation involving US patent 7305585 across patent litigation databases.

Let me search more targeted sources for this specific patent.

No litigation has appeared yet. Let me check the Unified Patents litigation portal and other sources directly.

Litigation Search: US Patent 7,305,585

Bottom line: I found no evidence of any patent litigation involving US 7,305,585 ("Asynchronous and autonomous data replication," inventor Benoit Marchand, assignee eXludus Technologies Inc.).

What I searched

I ran targeted queries for the exact number 7305585 and for "7,305,585" combined with litigation/complaint terms, and for the assignee (eXludus Technologies) plus the inventor name. The results returned:

• The patent's own Google Patents family record (US20040034822A1, US7707457B2 as a continuation, plus continuation-in-part US20050060608A1 and later US20080222234A1).
• The patent's prosecution/administrative record (assignments, maintenance-fee events).
• No district court complaints, no ITC actions, no CAFC appeals, and no PTAB/IPR proceedings naming the '585 patent as an asserted or challenged patent.
• No hit on the Unified Patents litigation case list for this number.

Relevant facts from the record (which bear on why no litigation exists)

Caveats

• Patent litigation databases (Unified Patents portal, CourtListener/RECAP, Justia dockets, Docket Alarm) are the authoritative places to confirm this. My searches surfaced no case naming '585, but I cannot rule out a case that is (a) not indexed by the general web search, (b) filed in a state court, or (c) sealed.
• I made no attempt to broaden the search to similar numbers, and I did not substitute any other patent for 7305585.
• The patent lapsed in 2015 for nonpayment of maintenance fees, which is consistent with there being no active assertion campaign — but it does not by itself prove no earlier suit was ever filed.

Conclusion: Based on the searches performed, no litigation is known involving US patent 7,305,585. This is a negative finding, not a certification that no case exists; a direct query of the Unified Patents case list (portal.unifiedpatents.com/litigation/caselist), PACER, and Docket Alarm for the exact number 7305585 would be needed to state that with certainty.

Proceedings overview

Total AIA trial proceedings on US 7,305,585: 0. The structured USPTO Open Data Portal block in this prompt returns no IPR, PGR, CBM, or DER proceeding, and independent web searches against Google Patents, PTAB/PatentsView-style decision sources, Justia, and CourtListener surfaced no IPR, PGR, CBM, reexamination, or Federal Circuit appeal naming this patent — so the breakdown by status is 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution-denied. The bottom line for a defendant is not "the patent is hardened" — it is that no third party ever bothered to attack it at the Board, because it lapsed for failure to pay maintenance fees effective 2015-12-04 (USPTO Legal Events: LAPS "Lapse for failure to pay maintenance fees," FP "Lapsed due to failure to pay maintenance fee," effective 20151204). A demand letter citing US 7,305,585 is citing an unenforceable, expired patent.

No proceedings to report

There is no proceeding number, petitioner, panel, institution decision, final written decision, settlement, or appeal to summarize for US 7,305,585. I will not manufacture one. Everything below is the absence analysis plus three important non-events.

What was checked

• ODP "PTAB proceedings on file" block (canonical): empty for this patent.
• Web search for 7,305,585 / 7305585 + IPR / PTAB / reexamination: no hits on this patent.
• Web search for eXludus / assignee litigation: no results returned.

Critical caveat on my verification. I could not open the PTAB E2E docket for this patent or run a live ODP API query myself; my conclusion rests on the canonical structured block supplied in this prompt plus negative web-search results. That is strong but not the same as a direct E2E docket pull. If you need a belt-and-suspenders confirmation for a litigation hold file, run the patent number through PTAB E2E (https://ptacts.uspto.gov/) and the ODP proceedings endpoint directly.

Three near-miss patents you will find if you search carelessly — do not conflate them:

1. US 7,075,585 (CF Crespe / MaxLinear, IPR2014-00728, IPR2015-00592) — different patent entirely; claims 1 and 17 held unpatentable, Fed. Cir. affirmed in part and vacated/remanded as to dependent claims 4, 6–9, 21. Same last-four-digits pattern, unrelated art (broadband TV receivers).
2. US 8,305,585 (Sartorius; Gator Bio v. Sartorius, IPR2023-00215, institution denied 2023-06-20 under § 325(d)). Again a different patent.
3. The "Marchand" prior-art reference (WO 01/76154 A2) used repeatedly in the IXI Mobile / Apple / Samsung IPRs (e.g. IPR2015-01443 on US 7,295,532) is a different Marchand, cited against IXI's Bluetooth piconet patents. It is not inventor Benoît Marchand's replication patent and has nothing to do with US 7,305,585.

Strategic summary

Claim status. Because no AIA trial ever reached a final written decision, none of claims 1–28 has been canceled, narrowed, or adjudicated by the PTAB. Every claim — independent claims 1, 20, and 25 and their dependents — is formally UNTESTED at the Board. That is an unusual posture, and it is not good news for the patent owner: an untested claim set is only valuable if it can be asserted, and this one cannot.

The real story is the lapsed maintenance fee, not PTAB. The USPTO Legal Events show the 4-year fee paid 2011-05-11, a maintenance-fee reminder mailed 2015-07-17, and a lapse for failure to pay maintenance fees effective 2015-12-04, followed by a status of "patent expired due to nonpayment of maintenance fees under 37 CFR 1.362." Google Patents also labels the file "Expired - Fee Related" with an adjusted expiration of 2024-11-18 (the original 20-year term from the 2002-05-23 priority date, as adjusted). Either way, the patent ceased to be enforceable in 2015; the statutory window to petition for revival of an unintentionally lapsed patent (37 CFR 1.378) closed long ago. Practically: no infringement can accrue on this patent today, regardless of claim scope.

Estoppel landscape. There is no § 315(e)(2) estoppel because there was no IPR/PGR — no petitioner, no privies, no instituted grounds. If someone asserted this patent today (for pre-2015 conduct), a defendant has the entire universe of § 102/§ 103 art available, unconstrained by any Board record. Ironically, the best "invalidity" position is even simpler than an IPR: the patent is expired and unenforceable for pre-expiration conduct only, and even that conduct would need to predate 2015-12-04.

Pattern signals. No repeat petitioner, no parallel IPRs, no defensive aggregator (no Unified Patents proceeding), no PTAB appeal history, no Federal Circuit docket. The patent was never litigated to the point of attracting an AIA challenge. The family continuation US 7,707,457 ("Completing an interrupted data replication operation," filed 2007-10-28, granted 2010-04-27) is also flagged "Expired - Fee Related" on Google Patents — if a demand letter maps to the family, check that sibling's maintenance history too, but the same "no PTAB activity" conclusion applies.

Recommended next steps

• If you are a defendant now: do not spend IPR money. Pull the patent's maintenance-fee record from USPTO Patent Center and confirm the 2015-12-04 lapse. An expired-for-nonpayment patent supports a motion to dismiss (or at minimum a summary judgment of no recoverable damages for post-2015 conduct). There is no FWD to link to — the dispositive document is the Legal Events tab showing FP / "Lapsed due to failure to pay maintenance fee," effective 2015-12-04, which you can cite from https://patents.google.com/patent/US7305585/en (Legal Events) and corroborate on USPTO Patent Center.
• If you are evaluating the portfolio, not this patent: the Marchand/eXludus family includes US 2005/0060608 (throughput compute clusters), US 2005/0216910 (software installation modules), and US 2008/0222234 (virtual environment deployment). Those are separate applications with their own maintenance and enforcement histories — screen each one rather than assuming the US 7,305,585 lapse carries over.
• If a troll demands a license: the correct response is that the patent lapsed in 2015 and cannot be infringed. Absence of any PTAB activity here is not the usual "no one has tested it yet" signal — it is the signal that the patent was never worth anyone's filing fee and is now dead weight.
• If you want independent verification: query PTAB E2E at https://ptacts.uspto.gov/ by patent number 7305585 and the ODP AIA-proceedings endpoint; both should return zero rows. Any result you find there that contradicts this memo takes precedence — but I found none.

Confidence statement: High confidence that no AIA trial proceeding exists on US 7,305,585. Moderate-high confidence on the maintenance-fee lapse (it comes from Google Patents' rendering of USPTO Legal Events in the authoritative text supplied; a direct Patent Center pull would make it airtight). I did not find and therefore do not assert any petitioner, panel, ground, FWD, settlement, or CAFC docket for this patent.

I'll verify the assignment record and the assignee's corporate history against live sources before writing up.

Inventors

Unusual patterns: None found. There is exactly one named inventor on the '585 patent, and the entire family — including continuation US 7,707,457 B2 (granted 2010-04-27) and the CIP/related filings US 2005/0060608 A1 and US 2008/0222234 A1 — is credited to Marchand alone. No co-inventor departures, no inventor-side re-assignment of the type that often precedes a portfolio fire-sale. Marchand is still associated with the entity as founder through at least the 2010–2012 period (management profiles list CFO/CTO appointments at eXludus in July 2010). I found no evidence he transferred, licensed away, or abandoned the family; the only recorded conveyance is the inventor-to-company confirmation assignment below.

Caveat: incorporations data for eXludus disagree (2002, 2003, Feb 2004, and 2005 all appear across different sources). The 2002 date comes from a contemporaneous technical evaluation; commercial databases say 2003. This does not affect the ownership chain.

Original assignee

eXludus Technologies Inc., Montréal, Québec, Canada (U.S. operations office listed at 19925 Stevens Creek Blvd., Cupertino, CA). Recorded at the USPTO as owner at Reel 015932 / Frame 0498.

• Primary line of business: HPC/grid middleware. The company's core product was Replicator, a closed-source multicast/broadcast file-replication engine, later followed by MCOPt, a multi-core resource-optimization kernel module.
• Did it ship a product embodying the claims? Yes — well documented. The "Replicator" product implements the claim-1 architecture directly: the vendor's own materials describe "rapid, asynchronous provisioning of files to a cluster of machines," broadcast/multicast distribution to "tens of thousands of machines," and independent asynchronous results transfer — i.e., separation of the multicast transfer phase from the recovery/replication phase. The independent White Rose Grid / UK e-Science evaluation (Report UKeS-2007-04, tests dated 01/02/2007) benchmarked Replicator v2.0.1 against NFS and GPFS and confirmed the multicast scaling behavior the patent claims (report PDF). So this is an operating-company patent that was reduced to practice in a commercial product, not a paper patent.
• Current status: Not operating; dissolved. PitchBook lists eXludus as "Acquired/Merged," M&A completed 01-Jul-2014, acquired by Filytics (PitchBook profile); Crustdata lists the entity as "Dead." The Québec enterprise register shows the corporate registration "TECHNOLOGIES EX LUDUS INC." cancelled 11 December 2014 (registre file 1162013016).
• Bankruptcy? No evidence. No Chapter 7/11 petition, receivership, or insolvency filing surfaced, and no assignment-to-liquidator was recorded at the USPTO. The 2014 outcome appears to be an ordinary tuck-in sale/dissolution, not a court-supervised patent auction.

Note: I found no evidence the patents were included in the Filytics transaction or any other patent-level transfer. The absence gives the chain its shape — the patent died with its owner rather than being packaged for assertion.

Assignment timeline

The USPTO assignment record for US 7,305,585 contains exactly one recorded post-issuance/pre-issuance assignment. Everything else in the record is prosecution, publication, priority, and maintenance-fee activity — not a conveyance.

• 2005-02-23 (executed) / recorded 2005-04-22 — Reel 015932 / Frame 0498
  • Conveyance: Assignment (Assignment of Assignors' Interest)
  • Assignor: Benoit Marchand (named in the record as "MARCHAND, BENOIT")
  • Assignee: EXLUDUS TECHNOLOGIES INC., Canada (recorded owner name)
  • Correspondent: Not determinable from the sources available to me. The Google Patents legal-events extract I was given reproduces the reel/frame, execution date, recording date, owner name, and the free-format text ("ASSIGNMENT OF ASSIGNORS INTEREST; ASSIGNOR:MARCHAND, BENOIT"), but does not expose the correspondent of record (the filing attorney/agent). I attempted to retrieve the correspondent directly and could not; I am not going to invent one. The assignment document itself is publicly retrievable at Reel 015932 / Frame 0498 via the USPTO Assignment Center, and the correspondent name appears on that image.
  • Context: Confirmatory founder/inventor-to-company assignment. Executed ~21 months after the 2003-05-23 U.S. filing and ~33 months after the 2002-05-23 EP priority — the classic pattern for a startup that files early and paper the chain of title later, typically to satisfy a financing or customer due-diligence requirement. Consistent with the July 2005 seed round ($1.5M, Brightspark Ventures) reported for eXludus.

Non-conveyance events in the same record (for completeness, not assignments): application filed 2003-05-23; publication of US 2004/0034822 A1 on 2004-02-19; priority claims to US 10/893,752 (2004-07-16), US 11/067,458 (2005-02-24), US 11/926,064 (2007-10-28), US 12/045,165 (2008-03-10); grant 2007-12-04; 4-year maintenance fee paid 2011-05-11; maintenance-fee reminder 2015-07-17; lapse 2015-12-04 and expiration for non-payment under 37 C.F.R. § 1.362 (final events 2015-12-31 / 2016-01-04 / 2016-01-26).

Bottom line for this section: one reel/frame, inventor → operating company, nothing downstream. That is itself the substantive finding.

Timeline diagram

timeline
    title Ownership of US 7305585
    2002 : EP priority application filed
    2003 : US application filed
    2005 : Assignment to eXludus recorded
    2007 : Patent granted
    2014 : eXludus acquired by Filytics
         : Quebec registration cancelled
    2015 : Maintenance fee lapse
    2016 : Patent expired

NPE / troll-pattern signals

Additional negative signal worth recording: the patent lapsed for non-payment of the 8-year maintenance fee (events of 2015-12-04 through 2016-01-26; adjusted expiration listed 2024-11-18). An NPE that had acquired this asset would not have let the maintenance window close — this is affirmative evidence against assertion-oriented ownership, independent of the absence of litigation.

Verdict

Insufficient data (only the original assignment).

The USPTO record contains a single conveyance — Benoit Marchand to eXludus Technologies Inc., executed 2005-02-23, recorded 2005-04-22 at Reel 015932 / Frame 0498 — and no downstream transfer of any kind. There are therefore no NPE signals to score and no asserter chain to trace; the correct verdict on the rubric is "insufficient data," not "non-asserting," because the absence of a chain here reflects a one-assignment record rather than an affirmative defensive-aggregator acquisition. Substantively, however, every available datapoint points the other way from an NPE: the assignee was a real operating company that shipped the claimed subject matter in its Replicator product (benchmarked on the White Rose Grid in Report UKeS-2007-04), the inventor is the company's own founder, no assignment was ever recorded to any licensing entity, no litigation naming the patent exists, and the patent was ultimately allowed to expire for non-payment of maintenance fees in December 2015. The one gap that must be closed by hand is the correspondent of record for Reel 015932/Frame 0498, which I could not retrieve from available sources and which should be read off the assignment image.

Verification: USPTO Assignment Center — search patent number 7305585 → https://assignmentcenter.uspto.gov/ (mirror: https://assignment.uspto.gov/patent/index.html); ownership/legal-events corroboration at https://patents.google.com/patent/US7305585B2/en.

Cross-section consistency check: This analysis is consistent with, and adds the ownership explanation for, the earlier litigation finding of "no known litigation." One item to flag for the record: the litigation section cites the maintenance-fee lapse as "consistent with there being no active assertion campaign." The assignment record now makes the stronger version of that claim supportable — there was never a downstream assignee who could have asserted it, so the lapse is unsurprising rather than merely suggestive.

I'll verify the patent record and the key cited references before building the anticipation analysis.

Prior-Art Analysis — US 7,305,585 B2

Building on the previously generated Patent Summary and Litigation Summary (bibliographic data, claim set, family, expiry). Those sections established the record; this section adds the citation-by-citation prior-art mapping and does not repeat the claim text or the status/term analysis.

0. Source verification and scope caveat

I could not open USPTO PatentCenter / Patent Examination Data System (PEDS/PAIR) directly from this session, so I did not retrieve the examiner's file-wrapper citations (Form PTO-892 / "Notice of References Cited"). What I can do — and have done — is work from the authoritative full patent text supplied to me, whose "Patent Citations (30)" block is the face-of-patent reference list, corroborated against the granted PDF (US7305585.pdf) and Justia's rendering of the printed claims. The identifier 7305585 was searched literally; I returned no other patent numbers as substitutes (the coincidental hits the previous sections flagged — US 7,075,585, US 6,730,585, US 7,290,864, the "Trimoprostil" registry hit — remain excluded).

• Full text / face-of-patent citations: https://patents.google.com/patent/US7305585/en
• Granted PDF (contains the front-page references): https://patentimages.storage.googleapis.com/92/19/5a/13a61e7fdee488/US7305585.pdf
• Justia (printed claim set): https://patents.justia.com/patent/7305585

Where a reference's content is described only from its title, assignee and field, I flag it as "(title-level description — not independently verified against full text)." I have verified full text only for the references I searched (noted inline).

1. The § 102 framework that governs these references — important

US 7,305,585 is a pre-AIA patent. Its priority date is 2002-05-23 (EP 02011310.6), so pre-AIA 35 U.S.C. § 102(a)/(b)/(e) applies, not the AIA § 102(a)(1)/(2) regime.

Two consequences drive the whole analysis:

1. Publication/issue date ≠ prior-art date for patents and published applications. Under pre-AIA § 102(e), a US patent or US published application is prior art as of its US filing date, provided it is "by another" and was filed before the applicant's invention date. Nearly every one of the 30 cited references was filed well before 2002-05-23 (most in 1993–2000), so they are § 102(e) art even though many issued in 2002–2004, i.e. after the priority date. This is "secret prior art" as to the post-priority-issuing ones (e.g., US 6,505,253; US 6,522,650; US 6,557,111; US 6,567,929; US 6,704,842).
2. § 102 anticipation is single-reference and requires every limitation. A reference "anticipates" a claim only if it discloses all limitations arranged as in the claim. Almost everything below is, at most, element-level evidence — strong for § 103 (alone or in combination) but not dispositive § 102 art. I say so explicitly per row rather than overclaiming.

Claim 1 = the gating test. Independent claims 1, 20 and 25 each require both:

• (A) transfer of data using a broadcast transfer protocol, (B) the protocol being sessionless; and
• (C) symmetric recovery of corrupted/missing data from any one of the networked computing devices, (D) the recovery being fully asynchronous and autonomous relative to the transfer.

A reference must hit A+B+C+D to anticipate claim 1 (and derivatively 2–19; similarly for 20–24 and 25–28).

2. The 30 face-of-patent citations — tiered

Effective prior-art date = US filing date (§ 102(e)); the "issued/published" column is shown for citation completeness and marked where it post-dates the 2002-05-23 priority.

Tier 1 — Reliable multicast + error repair (the art that touches limitation C)

Tier 2 — Broadcast/multicast data distribution, incl. late-join (limitations A/B)

Tier 3 — Replication / differential update / software distribution (dependent-claim material)

Tier 4 — Fault-tolerant multiprocessor / distributed agreement (background for the fault-tolerance theme)

Tier-4 references (24–30) are fault-tolerance/agreement background; none addresses packetized broadcast file transfer or peer data repair, and none individually touches limitations (A)–(D). They cannot anticipate claims 1/20/25 or the dependent claims. They are, at most, general § 103 context for the "fault-tolerant" framing, and the applicant cited them in that spirit.

3. Element-by-element — the genuinely closest art

3.1 Non-patent literature: the SRM / local-recovery family (the real § 102/§ 103 threat)

Although the task asks about patent citations, the closest prior art to independent claims 1/20/25 is in the patent's Non-Patent Citations (19) — specifically the scalable-reliable-multicast / receiver-based local recovery literature:

Assessment: This NPL family discloses elements (A)+(C)+(D) for a multicast data-distribution context and would be the examiner's or a challenger's primary § 102/§ 103 springboard. The only thing keeping claim 1 alive against it is limitation (B) — "the transfer protocol is sessionless" (no registration/handshake, per the specification's own definition of connectionless / redundant packets). If "sessionless" is construed narrowly (application-layer unawareness of peers, as the related '234 application defines it: "an application layer module need not be aware of its peer(s) presence to operate"), SRM's session model does not read on it.

3.2 US 6,256,673 (Intel) — best patent art on limitations (A)+(B)

Verified full text. The abstract and claims 9–10, 18, 22 disclose: cyclic multicast/broadcast of a packetized image file, with clients joining "at any moment in time without the need to synchronize with the central server's transmission," and a client that joins mid-cycle "download any missing portion of said full image file during said next … transmission cycle."

• Reads on limitation (A) (broadcast/multicast) and strongly supports limitation (B) (no registration/synchronization → sessionless from the receiver's standpoint).
• Reads on the subject matter of claims 2 and 9 (transfer of previously-transferred content to a later-introduced node; subsets/packets of a file).
• Fails limitation (C): repair is by the server re-broadcasting the next cycle, not "from any one of the networked computing devices" symmetrically. Conclusion: does not anticipate claim 1/20/25. Its real value is as the (A)+(B) half of a § 103 combination with a local-recovery reference (Tier 1 or the SRM NPL) supplying (C)+(D).

3.3 US 6,567,929 / US 6,501,763 (AT&T) — recipient/originator-initiated multicast repair

These disclose automatic repair of IP multicast sessions. The recipient-initiated variant (#10) is the closer one: it has (A) multicast, (C) repair of missing multicast data, and (D) repair decoupled from initial transmission. But repair is performed by a network-based service / repair server(s), not "symmetrically … from any one of the networked computing devices," and the scheme is built on IP multicast sessions (defeating (B)). So: no full anticipation of claim 1/20/25; strong element-level art for (C)+(D) and for dependent claims 2, 3, 14.

3.4 US 6,112,323 (Microsoft) / US 6,415,312 (IBM) — scalable reliable multicast repair

Both address scalable, low-overhead multicast repair (suppression of redundant feedback/repair). They supply (A)+(C)+(D) and are relevant to claims 14 and 19 (point-to-point repair exchange; scale-limited feedback traffic), but neither teaches a sessionless broadcast protocol nor peer-symmetric repair sourcing → not anticipatory of claims 1/20/25.

4. Bottom line

1. No single patent of record anticipates independent claim 1, 20 or 25. Every candidate misses at least one of (A)–(D):
   • Tier-1 reliable-multicast patents and the SRM NPL supply (A)+(C)+(D) but are session-based → miss (B).
   • US 6,256,673 (Intel) supplies (A)+(B) and late-join recovery, but its recovery is server-rebroadcast, not (C) symmetric peer recovery.
   • The AT&T repair patents supply (A)+(C)+(D) but via a network repair service, not symmetric peer sourcing, and are session-based.
2. The most relevant prior art overall is US 6,256,673 (patent art) in combination with the Floyd/SRM and Kasera/Liu local-recovery non-patent literature — but that is a § 103 obviousness case, not § 102 anticipation.
3. Most on-point for the dependent claims: US 6,557,111 (Xerox) for claims 9–12 (delta/update propagation); US 6,567,929 (AT&T) for claims 2, 3, 14; US 6,505,253 (Sun) and US 6,522,650 (Intellon) for claim 15 (traffic throttling/partial ARQ); the Microsoft installer/distribution family (US 6,418,554; US 2003/0145317; the '077 family) for claims 5, 16–18; US 6,704,842 (HP) for claim 14.
4. Tier-4 references (US 3,905,023; 4,130,865; 4,228,496; 4,412,281; 4,569,015; 4,644,542; 4,718,002) are fault-tolerance background only and cannot anticipate any claim.

5. Caveats (stated explicitly)

• Not a file-wrapper analysis. The list analyzed is the face-of-patent "Patent Citations (30)." The examiner may have applied a different subset and/or additional art in the PTO-892; only a PatentCenter/PEDS pull (or the PDF's front page) certifies which references the examiner actually relied on and how (anticipation vs. § 103 combination vs. background).
• Descriptions flagged "(title-level description)" are inferred from the reference's title, assignee, filing year and field; I verified the full text only for US 6,256,673 and (partially, via Google Patents) the AT&T and IEEE/SRM materials. Full-text retrieval of the Tier-1/Tier-3 patents in the USPTO/Google Patents image store is necessary before any § 102 conclusion is litigated.
• Claim-20 wording. The previously flagged abridgement of claim 20 in one snippet persists; the operative limitations are as stated, but use the printed claim set for exact language.
• Legal standard reminder. "Potentially anticipates" in the tables means discloses subject matter reading on the listed claim's limitations (element-level evidence). Only a single reference disclosing all limitations of a claim is a § 102 anticipation; everything else here is § 103-supported reasoning.
• The 2015 maintenance-fee lapse (from the Litigation Summary) affects enforceability timing but not the prior-art analysis above, which turns on the 2002-05-23 priority and the 2003-05-23 filing.

I'll ground this obviousness analysis in the actual prior-art references cited on the face of US 7,305,585 and in its background, then map them element-by-element.

Obviousness Analysis — US 7,305,585 B2 under 35 U.S.C. § 103

Reading note. This analysis builds on the earlier sections (bibliographic, litigation, PTAB, strategic) and does not repeat them. One cross-reference worth restating because it colors the whole exercise: the patent lapsed for non-payment of maintenance fees effective 2015-12-04, so this § 103 analysis is prospective/hypothetical for pre-2015 conduct only. It is still the right analysis to run — a validity opinion on an expired-but-once-assertable claim is exactly what a file-history (or a "what if" defensive) analysis needs.

Governing law. Priority is 2002-05-23, so pre-AIA § 103 applies. The controlling framework is Graham v. John Deere (scope/content of prior art; differences; PHOSITA level; secondary considerations) as refined by KSR Int'l v. Teleflex (motivation may come from the prior art, the nature of the problem, or common sense; predictable variation of a known technique is obvious; a "finite number of identified, predictable solutions" yields obviousness). I apply the patent's own cited references as the primary corpus, because those are the references the examiner and the applicant put on the table.

1. Level of ordinary skill (PHOSITA)

A bachelor's degree in CS/EE (or equivalent) plus 2–4 years of experience in distributed systems / reliable-multicast network protocol design, or a master's degree plus 1–2 years. By the 2002 priority date, the PHOSITA was thoroughly familiar with IP multicast (RFC 1112, Deering — cited), SRM/receiver-based repair (Floyd — cited), sender- vs. receiver-initiated error control (Towsley — cited), and local-recovery classifications (Kasera — cited). This is a mature art; the PHOSITA is skilled at combining a best-efforts multicast distribution mechanism with a separate repair/replication mechanism.

2. Claim 1 — element mapping

Claim 1 has only two real limitations. I take them in the order they appear.

Ground 1 (primary): Intel '673 + AT&T '929 + Kasera (self-organizing receiver-based recovery), optionally + Floyd SRM

Intel '673 discloses a multicast/broadcast file distribution in which a late-joining client machine (i) begins receiving mid-stream, (ii) identifies missing portions, and (iii) obtains those missing portions in a subsequent cycle — with no requirement to synchronize with the server. That reads on element (1a) and much of (1b).

AT&T '929 discloses recipient-initiated detection of missing packets and automatic repair — the repair is initiated by the receiver, not gated by the sender's multicast transmission. That reads on the "autonomous/asynchronous" aspect of (1d).

Kasera et al. (the patent's own NPL reference) discloses the self-organizing receiver-based protocol in which no repair server and no sender participation is required — a receiver dynamically selects another receiver to supply the repair, with randomized NAK suppression. That reads on (1c) — "from any one of the networked computing devices" — and supplies the "autonomous" quality of (1d).

Why the PHOSITA would combine them (KSR rationales). The motivation is stated in the references themselves: Towsley (cited) expressly reports that receiver-initiated error control provides substantially higher throughput than sender-initiated control; Kasera expressly reports that local/receiver-based recovery yields "significantly higher protocol throughput and lower bandwidth usage." The patent's own Background concedes that centralized multicast recovery "imposes an increased workload on the master file server node and eventually will limit scalability." The person of ordinary skill, facing precisely that scalability problem, had a strong, explicit incentive to move repair off the sender — the very thing Kasera and SRM teach. Under KSR, the combination is a predictable use of a known technique to solve a known problem.

Ground 2 (alternate): Xerox '111 alone or as the secondary reference

US 6,557,111 is arguably the single most damaging reference against claim 1 because it teaches, in one document, the exact two-phase architecture:

• a multicast update facility that propagates updates "from an originating replica source … to all replicas … at a single time using a best-efforts design" — element (1a)/(1b); and
• an epidemic (pair-wise) update facility that "send[s] updates by pair-wise interaction to non-updated neighboring replicas" — element (1c)/(1d), i.e., repair supplied by any peer, not the source.

The reference even says the two facilities are "configured … to operate in a parallel non-overlapping manner," which is a direct teaching of separation of the distribution phase from the repair phase — the asserted point of novelty in the '585 specification. A PHOSITA combining '111 with any reliable-broadcast reference (Intellon '650; Hughes '467) reaches claim 1.

The applicant's likely rebuttal, and its weakness

The '585 specification distinguishes prior art on three grounds: (i) registration requirement, (ii) recovery concurrent with the multicast phase, and (iii) no survival past master failure. Element (1b) ("sessionless") is the applicant's strongest hook. But Intel '673 was specifically designed to eliminate the registration/synchronization requirement, and Floyd SRM is a "light-weight session" framework in which a late member joins with no group-wide agreement. On the claim language as written — which says "broadcast transfer protocol … sessionless" and "recovery … asynchronous and autonomous relative to the transfer" — the reference combination meets it. The "fully asynchronous" qualifier is a manner-of-operation recital that SRM/Kasera satisfy (receiver-scheduled, randomized, suppression-driven repair) and that Xerox '111 satisfies even more squarely (parallel, non-overlapping pair-wise anti-entropy).

3. Dependent claims 2–19

Claims 4–12 largely recite well-understood replication life-cycle management (aging-out, garbage collection, delta/anti-entropy sync) that the cited references already teach in substance. Claims 13–18 are engineering refinements with explicit counterparts in the cited art.

4. The one non-trivial limitation: claims 19 / 21 ("adaptive distributed selection mechanism")

This is the only place the cited art is thin. The '585 specification describes the selection scheme as multicast of a random number and a modulus number, candidates applying the modulus and returning a second random number only on a match, with the modulus adapted based on the prior round's respondent count (initialized to "1" to force universal response).

The patent's cited election art — IBM US 4,569,015 (multiprocessor agreement), Tandem US 4,718,002, BBN US 4,130,865 — is the N×N agreement-based election the specification criticizes. So the specific modulus-matching scheme is not squarely disclosed by the references the examiner cited.

That said, the claim does not recite the modulus scheme. Claim 19/21 recites "generat[ing] constant and configurable network traffic using an adaptive distributed selection mechanism" — functional language that a PHOSITA would read onto:

• Floyd SRM's randomized, adaptive timers — "an adaptive algorithm that uses the results of previous loss recovery events to adapt the control parameters used for future loss recovery," with request/repair timers "set to a random number in intervals"; and
• Kasera's self-organizing receiver selection ("dynamically selects a receiver … to supply the repair").

Under the broadest reasonable construction (and, in litigation, under Phillips), a trier of fact could find claims 19/21 obvious over SRM + Kasera, on the rationale that randomized delay tuning to control aggregate control-message traffic is a known technique and the claimed "constant and configurable network traffic" is its predictable result. If, however, the patent owner successfully limits this claim to the modulus-and-match variant, then claims 19 and 21 are the strongest in the set and would require additional art outside the cited corpus (randomized/probabilistic leader-election and quorum literature). I flag this as the one point where I cannot close the § 103 case on the cited references alone.

5. Apparatus and CRM claims (20–28)

Claims 20 and 25 are the same two limitations as claim 1 in "configured to" / CRM form; they rise and fall with claim 1. Claims 21 (adaptive selection) carries the § 4 caveat; claims 22–24 and 26–28 merely recite node types (personal computer, PDA, cellular phone). Node-type recitations are quintessential non-limiting implementation details — the patent's own specification lists "a personal computer, a cellular phone, a PDA" as interchangeable examples of the claimed "computing device," which is an admission that the node type is not inventive. These add nothing to patentability.

6. Secondary considerations

None is apparent from the record. There is no evidence of unexpected results (the two-phase architecture delivers the scalability its references predicted); no nexus evidence of commercial success tied to the claims; and the patent's weak enforcement history (lapse in 2015; no PTAB or judicial challenge ever filed) undercuts any argument of industry praise or copying. Under KSR, absent a nexus, these factors do not tip the balance.

7. Bottom line

• Claims 1, 20, and 25 are likely obvious over US 6,256,673 + AT&T US 6,567,929 + Kasera/Kurose/Towsley (self-organizing receiver-based local recovery), and are arguably obvious over Xerox US 6,557,111 alone or with SRM (Floyd). The claimed "separation of transfer and recovery phases" — the heart of the asserted invention — is expressly taught by Xerox '111's parallel multicast + epidemic pair-wise design.
• Claims 2–18, 22–24, 26–28 are likely obvious over the combinations in the § 3 table.
• Claims 19 and 21 are the only genuine question. They are probably obvious over SRM's adaptive randomized timers + Kasera under a broad construction, but the modulus-match implementation is not disclosed by the cited art and would need additional references. This is the limitation a validity challenger should expect to litigate, and the limitation the patent owner should defend.
• Strongest defensive posture overall: the patent is expired and unenforceable for post-2015 conduct (per the earlier sections), so the § 103 analysis above is principally useful for attacking pre-2015 conduct or for portfolio/prior-art purposes.

Caveats and verification status

1. Grounding is on the patent's own cited art. I mapped every limitation above to references appearing in the "Citations" and "Non-Patent Citations" blocks of US 7,305,585 (Google Patents, https://patents.google.com/patent/US7305585/en). Two references surfaced in search but are not on the patent's face (US 5,805,825, Intel data carousel; US 8,305,585, Sartorius) — the first is used only as corroboration of the obsoleting/life-span feature and is flagged as supplementary; the second is unrelated and must not be conflated.
2. I did not obtain full printed claim charts for AT&T '929, Intellon '650, Hughes '467, or Compaq '059 — the mapping to those references is based on their titles/abstracts and, where available, quoted specification text. A final invalidity contention would require column/line pin cites to each.
3. Construction sensitivity. The outcome hinges on the breadth of "sessionless," "fully asynchronous and autonomous," and "adaptive distributed selection mechanism." Broad constructions make all claims obvious; narrow constructions (especially as to the modulus scheme) leave claims 19/21 intact. I state this explicitly rather than asserting a single result.
4. No contradiction with the earlier sections was found. The expiration/lapse facts in the prior sections are consistent with the record here; the only addition is the substantive § 103 mapping.