You forgot to mention heredocs! =)

My (personal and improvised) formula is:

• do you want variable interpolation?

  -> yes: " double $quote "

  -> no: ' single quote '

• can it get complex (embedded quotes, some multiline, other language (e.g HTML) )?

  -> yes: use the according quote-like-operators (q or qq) with convenient q!delimiter!.

• too complex to rule out embedded delimiters or too long to easily spot the end-delimiter?

  -> yes: use <<heredocs (i.e. multi-char-delimiter) or the <DATA>-section

• varying sources, volatile, scalable, multiuser, no control whatsoever

  -> yes: use files or a DB and/or a template system

qw is a special case of q which automatically splits on whitespaces.

Regexes and system-calls are different beasts, but rules are similar, cause they operate on interpreted strings as input. (NB s/// and m// allow other delimiters, too!)

quotemeta is very good if you can't control an "injected" string in an "interpreted" context like in a regex or with string-eval

As you can see it's mainly a compromise between readability, maintainability and avoiding delimiter-problems.

YMMV, readability is often a matter of taste.

It also depends on number of collaborators, performance, scalability, the expected future of the project, and so on...