unsup/g728-vt: code is not complete

[dennisguse]

The code of g728-vt is not complete and thus cannot be compiled.

[simaocampos]

The code in this unsupported module is a historic reference to the early G.728 verification process. G.728 was the first standard to use floating point arithmetic but still described as pseudo-code. G.728 Appendix I provides some tools for helping implementers check their code; being floating point, performance criteria involved SNR of the codec output versus reference ones, tools which were made available only in Fortran 77 (VAX VMS) and as executables under MS-DOS. (Finn Tore's work in early 1990's.) This is available within G.728 itself, not sure the executables will run. The code in g728-vt was a step towards having the verification tools in C and then as a step towards having a usable G.728 implementation available in the STL. But there were no volunteers to work with the code, so it was preserved for "future reference". My recollection is that it was generated from a Fortran-to-C translation but it was not fully operational. Some of the code (used to) compile and run. Now I question whether the code is still useful, since we have a G.728 implementation in the library, donated by AT&T several years after the codec was standardized (early 2000's if I recall correctly).

[ludomal]

Shall we drop it from the repository?

[simaocampos]

The background on this is that the code therein is an attempt to address a problem we have in G.728 Annex A, on testing G.728 implementations (done with test vectors). The Annex in G.728 refer to an executable for MS-DOS and some platform-specific Fortran code, for testing implementation-generated files against provided test vectors. As it can be imagined, using this executable is a complicated issue, and this package is the closest we got to a C implementation of that piece of code. Main problem was lack of volunteers to follow up the work. The researcher that did the code changed fields shortly after releasing the Fortran code. The current C code is almost the direct Fortran-to-C conversion of the code, that I generated back in the early 1990s but has not been worked on.

…

-- Simão