Regarding my previous post I’d like to add something more:
1)A printer has its own “margins” (non-printable portions of a page). This cannot be foreseen. I believe, that a “smart” PDF-application is capable of detecting
the bbox of all “printable” elements in a PDF, and thus be able to fit that bbox onto a piece of paper of perfectly chosen/best fitted size.
2)When trying to produce true-scale PDF-maps (with margins), pre-calculations and pre-processing have to performed up front. This, in order to achieve a
desired print quality of all raster and vector layers (DPI).
What I am saying is, that a producer of a true-scale PDF-map needs to be aware of lots of the aspects of the to-be generated PDF.
Thus, in my point of view, the input to gdal_translate need not to be easily understood (but documentation serves well