Dear Laurent,

"LS" == Laurent Siebenmann writes:

 LS> ** << Every letter should be input by TeX with category "letter"
 LS> (i.e. category code 11). >>

 LS> LaTeX does not accept this and currently inputs most cyrillic
 LS> characters with category "active" (=13).

i think that's not fully correct: LaTeX recommends one to use inputenc
package for processing texts with 8-bit input encodings; but LaTeX
will still work fine if one will not use inputenc; then one has to
ensure to use an alternative re-encoding mechanism (such as TCX or
TCP).

 LS> Malyshev's various TeX systems have all permitted users to
 LS> respect this tenet if they wish.

The same is valid for standard LaTeX's approach which permits users to
respect that tenet if they wish. On the other hand, inputenc has some
advantages like independence from non-standard extensions (TCX/TCP),
more accurate/full support for input encodings (inachievable with
TCX/TCP -- i mean constructing accentsd glyphs on the fly, and taking
them from other fonts, etc), and support for Unicode bookmarks in PDFs.

These and some other advantages explain why inputenc is a recommended
approach, and why TCX/TCP are not recommended.