Two questions about the project's provenance: why the original binary was not decompiled, and where the name comes from.
Back to the README.
Tensor.exe is a 208 KB 16-bit MS-DOS binary with 5252 relocations, no symbol
table, probably Turbo Pascal, with overlays and most likely software floating
point. It will not run on 64-bit Windows, which dropped NTVDM. Decompiling it is
not a viable route, for the usual reasons: compilation discarded the names, the
types and the structure, and 16-bit segmented addressing makes pointers
impossible to resolve statically.
The algorithm, on the other hand, is fully published:
- Angelier, J. (1990) Inversion of field data in fault tectonics to obtain the regional stress. III. A new rapid direct inversion method by analytical means. Geophys. J. Int. 103, 363-376.
- Angelier, J. (1984) Tectonic analysis of fault slip data sets. J. Geophys. Res. 89(B7), 5835-5848.
- Angelier, J. (1994) Fault slip analysis and palaeostress reconstruction. In: Hancock (ed.) Continental Deformation, ch. 4.
So the work went into reading the papers, and into measuring the original's own output files for everything the papers do not state.
Angelier's papers never name a program: the 1984, 1989 and 1990 texts speak only of "the new direct inversion method". The binary names itself, in every INFO1 it wrote:
Progr. TENSOR, 1975-1991, version 5.45, jan91
Copyright 1987,1988,1989,1990,1991 J. Angelier
*** DATA BASE FOR TECTONIC ORIENTATIONS "TECTOR" ***
"TENSOR" would have been the natural tribute, but the name is taken twice
over: in the palaeostress community it now means Damien Delvaux's unrelated
TENSOR / Win-Tensor program (Delvaux & Sperner 2003), and on PyPI pytensor
is PyMC's actively maintained array library, so the import name would collide
with real installations. TECTOR is the other name on that banner, it is
uniquely Angelier's, and nothing else uses it. Hence pyTECTOR.