Writing tar in PROLOG sounds "cool", a neat hack, but it is as practical as writing the Linux kernel in COBOL. Possible, perhaps, but not the best idea.

It's hard enough to write things in PROLOG that _are_ ideally suited for the language - like natural language parsers (they used to be LISP or PROLOG, but these days they are all in C++ or FORTRAN wrapped in Python so no-one gets hurt; why? Because of efficiency and numeric libraries required - modern parsers for human languages implement probabilistic or neural models, no longer symbol manipulation and pattern matching).

EDIT: I'm not a PROLOG hater - in fact, enjoyed reading the article (and I will read sequels on implementing "mount" in RPG3, "htop" in COMAL etc.) and my friends include some PROLOG lovers. I also find it important and stimulating to learn about another paradigm, and logical programming is an important one for many reasons. This rant is more about using it on the specific task of implementing "tar", for which it is a poor fit, as the article's PS shows.