History
From TNT
Briefly, TNT is the last link in the chain of different phylogenetic programs, such as Hennig86, Pee-Wee, NONA, and Winclada.
For more information about TNT or Cladistics, visit the Willi Hennig Society website.
The idea of the TNT project started in 1997, during the Washington Hennig meeting. At the time, the program Nona (by P. Goloboff) had some advantages over PAUP*, and WinClada (by Kevin Nixon) provided a nice user interface and other facilities. Steve Farris, who had released Hennig86 some years before (1988), approached Goloboff and Nixon with the proposal to join forces and create a program which could supersede all the others.
Next year, after the Sao Paulo Hennig meeting (Oct. 1998), Farris and Nixon visited Tucumán, with the idea of actually beginning work on the project. Kevin could stay only a couple days, due to health reasons. Farris and Goloboff then proceeded to discuss the general layout of the program, and Farris wrote the first lines of code of the new program (command interpreter, error handler, memory manager, etc.). Subsequently, Goloboff continued work, adding routines to read data, plot trees, consense, map characters, parse scripts, etc. No search routines had yet been incorporated into TNT. Nixon had just discovered the ratchet; next year (early 1999) Goloboff began exploring alternative algorithms (inspired mostly in the success story of the ratchet), and came up with sectorial searches, a refined tree-fuser (earlier attempts by Goloboff had been less successful, but in discussion with Nixon during his brief visit previous October, the problem with those attempts became evident --and thus could be corrected), and tree-drifting.
During late 1999, Kevin Nixon (already recovered from his health problems) made a new trip to Tucumán, wrote the first lines of code for the Windows interface of TNT, and taught Goloboff the basics of programming Windows interfaces. After Kevin left, Goloboff continued adding functionality to the menus, and during the 2000 Hennig meeting (in Leiden) the first public presentation of TNT was made. The program (still in version 0.2 or so) did its searches only for equally-weighted characters, and calling other programs as sub-processes (these programs were modifications of the quick-consensus estimators of Goloboff and Farris, 2001).
Finally, in 2003, Goloboff managed to find the time and energy to write all search routines for TNT from scratch (about 15,000 lines of code), and presented version 0.8 --close to fully functional-- at the NY Hennig meeting. At this stage, the program (with 45,000 or 50,000 lines of code) already had incorporated the possibility of using different weighting schemes (including Sankoff optimization), as well as implied weighting.
The most important subsequent additions to the program have been the addition of continuous characters (early 2004), parallel routines via PVM (late 2004), auto-weighted optimization (mid 2006, formerly implemented only in SL, an experimental program, slower and less flexible than TNT), the possibility to run on 64-bit machines and run data sets for any number of taxa (late 2007), and more recently, landmark characters (i.e. 2D or 3D coordinate data). As of May 2009, the program comprises about 90,000 lines of code, and something like 6,000 to 8,000 hours/man of work (and I am now waiting for the painter).
