Language util name byte counting

Question

I'm trying to figure out how the byte counting works in "Sum the numbers on standard in". In one answer using the jq util counts the non-space bytes after the util name, but not the util or the spaces. Using similar reckoning, (i.e. don't count the util name), one could get a byte count of 0 from numsum.

Possibly the distinction is between programs which are also languages, (i.e. jq describes itself as a language, therefore don't count the language binary name), and utils that are not, (so do count the util name). If so, perhaps the question is what distinguishes a language from a util with a lot of switches.

Supposing a clear distinction exists between language executables and utils, there's then the question as to language executables used as utils, as in this answer where the pipe |bc is counted even though bc describes itself as a language.

Answer 1

Basically, the rule is that the name of an executable counts if you have to access that executable during the course of running the code, rather than it being the thing that runs the code in of itself. In the example of the bash + bc answer, bc is counted as 2 bytes because those two characters are part of the Bash program (it needs them to figure out that it has to call into bc rather than some other executable), rather than anything to do with bc itself. bash isn't counted, even though you'd likely need to type it to run the program, because the program's a Bash program.

The general rule is that you take the shortest method of running a program in your language as "+0" bytes of penalties. Anything you need to do to the command line beyond that gives you byte penalties.

For what it's worth, I believe this challenge has a valid 0-byte answer () in numsum, and a valid 6-byte answer (numsum) in Bash + numsum. (Note how those are different languages. The former isn't a programming language; atm we don't consider that disqualifying but views on that have gone back and forth over time.)