1957 IBM FORTRAN 'The Fortran Automatic Coding System for the IBM 704' — John W. Backus's Solo Proposal, a 10-Person Team's Implementation, Reference Manual Released 1956-10-15, and Compiler Shipped April 1957, Yet Google Patents inventor=Backus / IBM, Wikipedia EN John Backus, ACM Turing Award Laureate Page, and Britannica All Carry No Patent Number: 'Eligibility Wall' Excavation Log #2 (SW Subseries DB Form: Eligibility Wall #2 / Origin of US Software-Patent Eligibility's Pre-Gottschalk Era 1957)

Conclusion First

Day 25 establishes the SW subseries' third note. The subject is IBM FORTRAN, with John W. Backus's solo proposal in late 1953, the first reference manual published 1956-10-15, and the first compiler shipped April 1957.

To the point: no patent number for FORTRAN itself was found within today's verify scope. The Google Patents query inventor=Backus John+assignee=IBM+priority:1953-1960 returns 0 results. The ACM Turing Award Laureate Backus official page, Wikipedia EN John Backus, Britannica Backus, and the History of Information FORTRAN article all carry no FORTRAN-related patent reference. The 1953-1957 period predates any US judicial precedent on software-patent eligibility. Gottschalk v. Benson (1972) / Parker v. Flook (1978) / Diamond v. Diehr (1981) / State Street Bank (1998) / Alice Corp v. CLS Bank (2014) — these five Supreme Court / Federal Circuit decisions later cumulatively defined the 'pure-algorithm-invention eligibility' boundary. FORTRAN is a pure-software invention 15 years ahead of this case-law arc, and IBM's strategy at the time consisted of three pillars: 'early reference-manual publication (1956-10-15) + free distribution to customers + internal trade secrecy of source code.'

Combining Day 24 SW-005 HyperCard (1987 — eligibility-unsettled era right after Diamond v. Diehr 1981) and today's ep88 SW-002 FORTRAN (1957 — pre-judicial era, 15 years before Gottschalk v. Benson 1972), the SW subseries' 'eligibility wall' form resolves into two sub-forms: (a) pre-judicial era (pre-1972) / (b) eligibility-unsettled era (1972-1998). Today's continuing ep89 SW-003 BBN IMP (1969 ARPA contract, mandated public spec) and ep90 SW-004 Xerox PARC Smalltalk (1972 corporate lab, voluntary 4-licensee public release) add (c) public-disclosure mandated by government contract / (d) voluntary public disclosure as corporate strategy. Day 25's editing achieves a four-form line-up of the eligibility wall in a single session.

1. How the Subject Was Selected (Reproducible Pipeline)

[STEP 1] Building on Day 24's SW-005 HyperCard Eligibility Wall #1,
         confirm with Haruko the plan to excavate SW-002 (FORTRAN) /
         SW-003 (BBN IMP) / SW-004 (Smalltalk) sequentially in Day 25

[STEP 2] Web verification on FORTRAN-related patents
   - Google Patents xhr API direct query
     URL: patents.google.com/xhr/query?url=inventor%3D%22John%2BW%2BBackus%22
          %26assignee%3DIBM%26before%3Dpriority%3A19600101%26after%3Dpriority%3A19530101
     Result: total_num_results = 0
   - WebSearch ""John W. Backus" inventor US patent IBM"
     Result: only unrelated patents from a different John W. Backus
     (medical diagnostics, US6300075B1 etc.); no FORTRAN-inventor patents

[STEP 3] Secondary-source cross-check
   - Wikipedia EN Fortran: detailed development chronology (1953-1957),
     no patent number reference, no IBM IP-strategy reference
   - Wikipedia EN John Backus: career covers Speedcoding (1953) /
     FORTRAN (1954-) / BNF (1959 ALGOL) / FP language (1977 IBM Research),
     **zero patent references**
   - ACM Turing Award Laureate Backus official page (1977 award):
     'for profound, influential, and lasting contributions to the design
     of practical high-level programming systems, notably through his
     work on FORTRAN, and for publication of formal procedures for the
     specification of programming languages' — no patent reference
   - Britannica John Backus: career & awards, no patent reference
   - History of Information article confirms: 1956-10-15 first reference
     manual, April 1957 compiler shipment, 10-person team
     (Richard Goldberg, Sheldon F. Best, Harlan Herrick, Peter Sheridan,
      Roy Nutt, Robert Nelson, Irving Ziller, Harold Stern, Lois Haibt,
      David Sayre); no patent reference

[STEP 4] Confirm legal status of US software-patent eligibility in 1957
   - 1972 Gottschalk v. Benson Supreme Court: rejected a BCD-binary
     conversion algorithm patent application; first ruling that 'mental
     processes are outside 35 U.S.C. § 101 patentable subject matter'
   - Pre-1972: **no precedent existed at all**; USPTO examination
     guidelines were not established. As an examination practice,
     'an algorithm = a mathematical method = unpatentable' was the
     operating norm.
   - FORTRAN (1957) sits 15 years before the case law, in
     **the pre-judicial era**

[STEP 5] Conclude: 'No FORTRAN patent record found in today's verify
        scope.' Treat this as SW subseries' Eligibility Wall #2.

2. FORTRAN Development Chronology (Cross-Verified Across Secondary Sources)

This chronology is cross-verified by Wikipedia EN Fortran, History of Information, and the IBM official Backus page. No record of FORTRAN-related patent applications or grants exists in any secondary source within this 1953-1972 window.

3. Core: The Structure of 'Eligibility Wall (a) Pre-Judicial Era Form'

(a) US Software-Patent Status in 1957

In 1957 the United States had neither judicial precedent nor USPTO examination guidelines on software-patent eligibility. The first significant ruling on patent eligibility was the 1972 Gottschalk v. Benson Supreme Court decision, which rejected a patent application involving a BCD-to-pure-binary conversion algorithm and held that 'mental processes are outside the scope of 35 U.S.C. § 101.' That ruling came 15 years after FORTRAN compiler shipment.

Before that, US patent practice in the 1950s adhered to the 'mathematical method = unpatentable' doctrine inherited from the 17th-century English Statute of Monopolies. Pure-algorithm inventions were effectively rejected at examination as a customary operating norm. The core technologies of the FORTRAN compiler — parsing, register allocation, code generation, expression optimization — are all algorithmic inventions; under the prevailing operating norms it was likely difficult to draft viable claims around them.

(b) IBM's 1957 IP Strategy: 'Manual Disclosure + Free Distribution + Trade Secrecy' — Three Pillars

In a context where patenting was difficult, IBM defended FORTRAN with three pillars:

1. Early reference-manual publication (1956-10-15): about half a year ahead of compiler shipment, the grammar specification, built-in functions, and I/O formats were published. This established FORTRAN as a de facto standard and limited the risk of competing vendors building compatible compilers from a different syntactic basis.
2. Free distribution to IBM 704 customers: the FORTRAN compiler itself was given free to IBM 704 buyers. Positioning the compiler as 'an accessory of IBM 704 hardware, not an independent product' meant the cost was internalized into hardware pricing as a viable business model.
3. Source code as internal trade secret: the compiler's source code was not published; it was managed internally as IBM confidential. Optimization-algorithm details were partially disclosed in ACM papers (1957 SHARE Conference, etc.) while the full implementation was retained as IBM IP.

The same three-pillar pattern continues, in problem framing, through Microsoft MS-DOS / Windows in the 1980s, Sun Microsystems Java in the 1990s, Apple Swift in the 2010s, and OpenAI GPT-4 / Anthropic Claude in the 2020s.

(c) Backus's Long-Term Strategy: A Mathematician's 'Algorithms Are a Public Good' Disposition

Backus's personal pattern of invention also seems to point away from patenting. In 1959, during the design of ALGOL 60, Backus proposed BNF (Backus-Naur Form), publishing it via ACM Communications (no patent). In his 1977 ACM Turing Award lecture, Backus proposed the functional programming language FP and again published in ACM Communications (no patent). Across Backus's 30-year career, the consistent mode was 'publish papers + standardize languages' — and no personal patent record was discovered within today's verify scope.

This is consistent with the era's cultural premise that 'programming-language designers belong to the mathematicians' community.' Backus's contemporaries — McCarthy (LISP, 1958) / Hoare (QuickSort, 1959 / CSP, 1978) / Dijkstra (Shortest Path, 1956 / Structured Programming, 1968) / Knuth (TAOCP, 1968-) — each chose to publish core algorithms as papers and not pursue patents.

4. Modern Correspondence Table (with 4-Level Evaluation)

How to Read the Table

Across the four-level evaluation (Same / Similar / Metaphor / Forced), only 1 of 6 rows is 'Same' (the unchanged § 101 statutory wording), 4 are 'Similar,' and 1 is 'Metaphor'. The honest summary: the FORTRAN (1957) ↔ LLM (2026) correspondence holds at the level of problem framing, but diverges at the implementation level.

The claim 'FORTRAN is a precursor of LLMs' should be read as a precursor at the framing level (programming abstraction, high-level language, invention strategy under eligibility walls). At the implementation lineage level, the direct technical predecessors of ML/LLM are LISP (1958) → Smalltalk (1972) → Self (1987) → JavaScript (1995) → Python (1991), and LISP → Common Lisp → Scheme → ML (1973) → OCaml (1996) → Haskell (1990) → PyTorch / JAX (2016-). FORTRAN is the upstream parent of these flows but not a direct implementation predecessor of ML/LLM.

5. Pitfalls (5 Items)

1. 'No FORTRAN patent record' ≠ 'Backus tried and was rejected': today's verify scope only confirms 'no application record' was found; 'application-rejected records' would require USPTO Patent Center, which is dialog-UI mandatory and not amenable to curl-level automated verification. IBM internal application-evaluation records were also not retrieved. Three possibilities remain: 'no application was ever filed,' 'an application was filed but never published,' or 'an application exists but was not found in today's verify scope.'
2. 'IBM in 1957 didn't take software patents' ≠ 'IBM took no patents': IBM held many hardware patents in the 1950s — IBM 704 mainframe itself, magnetic-core memory, I/O channels are all patented. The narrower factual claim here is that the FORTRAN compiler software portion was not patented.
3. Cannot conclude 'Backus opposed patenting': the absence of Backus's personal patent record in today's verify scope does not entail that Backus was ideologically opposed to patenting. Whether the primary cause was philosophy, culture, practice, or legal context cannot be determined within today's verify scope.
4. Cannot conclude '1957 was a pre-judicial era so application was futile': even before Gottschalk v. Benson (1972), software-/algorithm-invention applications were technically possible, and a small number of 1960s application examples are recorded (Bernard Bull patents, etc.). The reason FORTRAN was not patented may be substantially due to IBM's strategic choice.
5. 'IBM's trade-secret strategy succeeded' is not unconditional: from the late 1960s, compatible compilers (Univac / Honeywell / CDC) emerged, and FORTRAN moved away from IBM-only dominance via de facto standardization. IBM's strategy was completed not in FORTRAN but in System/360 (1964) / OS/360 (1966); evaluating it on FORTRAN alone is suboptimal.

6. To Be Strictly Accurate (Mandatory 5 Items)

Confirmed Facts

• The chronology — late 1953 Backus solo proposal / Nov. 1954 draft spec / 1956-10-15 manual / April 1957 compiler shipment — is cross-verified across Wikipedia EN Fortran, History of Information, and IBM's official Backus page
• The 10-person team membership (Goldberg, Best, Herrick, Sheridan, Nutt, Nelson, Ziller, Stern, Haibt, Sayre) is verbatim-confirmed in Wikipedia EN Fortran: 'Backus' historic FORTRAN team consisted of programmers Richard Goldberg, Sheldon F. Best, Harlan Herrick, Peter Sheridan, Roy Nutt, Robert Nelson, Irving Ziller, Harold Stern, Lois Haibt, and David Sayre.'
• The Google Patents xhr API query (inventor=John W Backus + assignee=IBM + priority:1953-1960) returns 0 results, and WebSearch likewise yields no FORTRAN-related Backus patent
• The 1972 Gottschalk v. Benson decision being the first significant US software-patent-eligibility precedent is established legal history

Author's Interpretation

• The three-factor explanation — 'FORTRAN was not patented because of (1) eligibility wall, (2) trade-secret strategy, (3) Backus's mathematician culture' — is this author's interpretation; it has not been cross-checked with IBM internal records or Backus's personal correspondence
• The observation 'paper-publication + standardization was consistent across Backus's 30-year career' was extrapolated from the 2 documents (BNF paper 1959, Turing FP paper 1977); patent-record existence for Speedcoding (1953) / FP implementations (1977 IBM Research internal) was not separately verified
• The 'three-pillar 1957 IBM strategy' is an after-the-fact structural reconstruction; whether IBM management explicitly intended this triad at the time is unverified

Metaphors and Analogies

• The FORTRAN (1957) ↔ LLM (2026) correspondence distributes as 4 'Similar' + 1 'Same' + 1 'Metaphor'; the implementation lineage runs LISP → Smalltalk → ML → PyTorch, and FORTRAN is the upstream parent of those flows, not a direct implementation predecessor of ML/LLM
• The two-form distinction 'pre-judicial-era eligibility wall (FORTRAN) vs eligibility-unsettled-era wall (HyperCard)' is an after-the-fact structuring from Day 24 SW-005 and today's SW-002; experts may push back that 'the case-law periodization is too coarse'

Unverified

• Comprehensive USPTO Patent Center search across Backus-named, IBM-1953-1972 FORTRAN-related patents (dialog-UI mandatory, not amenable to automated curl)
• IBM legal/IP department's internal application-evaluation records (would require IBM Research Almaden archive access)
• Full text of Backus's 1957 SHARE Conference paper (not reachable within Web scope)
• Comprehensive 1960s US software-patent application list including Bernard Bull et al. examples
• Backus's personal correspondence / unpublished papers (Library of Congress Backus papers — catalog PDF reached, but main contents unread)

Where This Comparison Breaks

• The accurate phrasing is 'no record was found within today's verify scope,' not 'FORTRAN's patent absence is confirmed.' If USPTO Patent Center or IBM internal archives surface related patents, this note will be updated
• The 'pre-judicial-era eligibility wall' concept is a tentative framework abstracted from the 2-sample base of Day 24 SW-005 and today's SW-002; whether the same form applies to other pre-1972 software inventions (COBOL 1959 / LISP 1958 / ALGOL 1960) is unverified, and excavation of those neighbors is needed
• The 'Backus's mathematician-culture orientation' observation extrapolates from 2 published papers across a 30-year career; to conclude that Backus was ideologically anti-patent would require closer reading of his statements (Turing Lecture transcript, interview records)
• The 'IBM three-pillar strategy' is an after-the-fact reconstruction; post-1980, IBM pivoted to massive software-patent acquisition (IBM has held the top US-annual-patent-grant rank for 30 consecutive years), so the '1957 strategy may have been a contextual response, not IBM's enduring essence' caveat applies

7. SW Subseries DB Form: 'Eligibility Wall' #2 (Pre-Judicial Era Form)

Pairing Day 24 SW-005 HyperCard (1987 — eligibility-unsettled era right after Diamond v. Diehr 1981) with today's ep88 SW-002 FORTRAN (1957 — pre-judicial era 15 years before Gottschalk v. Benson 1972) raises the resolution of the SW subseries' 'eligibility wall' form into two sub-forms:

Today's continuing ep89 SW-003 BBN IMP (1969, ARPA government contract, public-spec mandated) and ep90 SW-004 Xerox PARC Smalltalk (1972, corporate lab, voluntary 4-licensee public release) add further sub-forms:

Day 25's single session yields a four-form line-up of the eligibility wall, structuring the SW subseries' 7-entry DB into a two-group contrast: (1) patents successfully claimed via display-system / cursor-control-device claims that include hardware elements (SW-001 Engelbart Mouse / SW-006 Atkinson Image Compression) vs (2) inventions where the eligibility wall blocked patenting (SW-002 FORTRAN / SW-003 BBN IMP / SW-004 Smalltalk / SW-005 HyperCard).

References

• Wikipedia EN Fortran: en.wikipedia.org/wiki/Fortran
• Wikipedia EN John Backus: en.wikipedia.org/wiki/John_Backus
• ACM Turing Award Laureate John Backus: amturing.acm.org/award_winners/backus_0703524.cfm
• History of Information 'John Backus & Team Develop FORTRAN': historyofinformation.com/detail.php?id=755
• IBM Official John Backus: ibm.com/history/john-backus
• IBM Official Fortran: ibm.com/history/fortran
• USPTO Patent Subject Matter Eligibility: uspto.gov/ip-policy/patent-policy/patent-subject-matter-eligibility
• Related episode: #87 SW-005 Apple HyperCard Eligibility Wall #1 (Day 24 memo)
• Related episode: #89 SW-003 BBN IMP Eligibility Wall #3 (today's memo)
• Related episode: #90 SW-004 Xerox PARC Smalltalk Eligibility Wall #4 (today's memo)