Many languages have flexible word order and the trend in language evolution is that word order tightens. Classical Latin, and I believe Greek, don’t have a defined word order, yet most modern languages have one albeit loosely enforced.
This is a piece of mythology that has been around in grammars of Greek and Latin probably since Priscian (or insert name of early Greek or Latin grammar).
Syntax is more than just word order. In fact, it would probably be better to talk about constituent order within phrases and clauses when talking about the ordering of elements. In English phrases, the head is usually on the left and the modifier on the right. The former is obligatory; the latter isn’t. Phrase order is typically SVO with adverbials (A) having a degree of mobility (which seems mostly peripheral at least in English; e.g. The student bought a book yesterday ~ Yesterday the student bought a book).
Because English lacks inflectional endings (outside personal pronouns) marking categories such as subject and object, the order of elements is vital for understanding the meaning. English speakers would probably interpret “A book bought the student” as “The student bought a book”, but regard the sentence as affected. Highly inflected languages themselves are not necessarily laissez faire with respect to constituent ordering. The sentences
discipulus emit librumdiscipulus librum emitlibrum discipulus emitlibrum emit discipulusemit discipulus librumemit librum discipulus
all convey the same fundamental information because the inflections indicate the subject (discipulus) and the object (librum). However, if you could show these sentences to a native speaker of 1st century Latin and ask them to rank the sentences according to their naturalness, there’d be one which was unexceptional (probably the second) after which the rest would deviate increasingly from the norm and might need to be placed in context for such an order to make any sense.
There is some variety in the ordering of constituents in inflected languages, but it’s not usually the quodlibet which the writer of the comment above appears to be assuming. And outside of configurational languages (e.g. Warlpiri; Hungarian) “loose enforcement” doesn’t happen.