Karl > LaTeXers - personally, I think it would be nice if the "primitive > requirements" became a permanent part of the LaTeX > documentation. (Apologies if they already there.) Maybe? any suggestion where that should be? I consider ltnews actually not a bad place especially as there is also https://www.latex-project.org/news/latex2e-news/ltnews.pdf which puts them all together. more for @Thierry ... It might be possible to make a distinction between primitives that are specific to pdf generation (because they are not relevant if dvi is produced (which we intend to support)) and those that are for general coding, e.g. \ifincsname, \expanded or \pdfstrcmp (which has nothing to do with "pdf" other than it was first introduced in pdftex) and a number of others. The general coding primitives are rather essential and for most of them there is no way to emulate in any way that still allows a somewhat reasonable performance even if technically TeX is Turing-complete. In other words it should be possible to drop the primitives which are really PDF output specific (if we list any of them as required) and require them only for engines/engine-modes that target PDF directly and make sure that the format doesn't complain if they aren't there, but there is not reasonable way going forward without the general coding ones that are now (fortunately) available in all major engines. frank