Some interesting points, but I think you've missed a trick with your point about MS Word - not only is it fairly trivial to linearize a tree structure (some XML, for example), but you gain a fairly significant speed advantage reading and writing binary data as compared to text (something which will matter in the commercial market place - let alone the fact that, as you mention, it can be difficult to reverse engineer a binary file if you don't know what you're looking for). It is for this reason that many popular file formats are available in both text and binary forms: for example(s) Maya's files (.mb and .ma), the well-known Stereolithography format (.stl can contain ascii or binary), Parasolid files (.x_t and .x_b), etc etc. - Another point worth bringing up (particularly in relation to the file types I've mentioned) is the ability to _choose_ between text or binary, which gives you, as a user of the file type, the option to inspect your output data as text or to load/save quickly as binary.

Yea really