Reinhard Kotucha wrote:

Nobody forces you to use double quotes. TeX still works as it did 20 years ago. Why do we need a TeX parser to process filenames?

We don't. But we do need to know how TeX operates, and in this context it is a many-layered operation. Scanning, parsing and tokenization all take place, as do execution and expansion, and with (I think) the sole exception of \lowercase and \uppercase (which are /horribly/ anomalous), each phase is quite distinct from the other. What you are suggesting (if I understand you correctly) would make \input even more anomalous than \...case, and that would (I believe) be a very retrograde step in the evolution of (derivates of) TeX.

What is the advantage if files are inaccessible?

They are not. It is not impossible (or even outrageously difficult) to construct a control sequence that will expand to "foo<space><space>bar"; what /cannot/ be achieved is for "foo<space><space>bar" to appear in the input stream with normal catcodes and for the two spaces to still be there after the scanner has performed its function. ** Phil.