RFC: "use re 'strict'"

I'm proposing two additions to the "use re" pragma.  One is covered 
here, and the other in a separate thread, to lessen confusion.

'strict' would mean that any regex compiled within its scope would be 
subjected to stricter checking than otherwise.  For example, currently,

$ perl -Dr -le 'qr/[z[:ascii]]/'
Compiling REx "[z[:ascii]]"
rarest char ] at 0
Final program:
    1: ANYOF[:[acisz] (12)
   12: EXACT <]> (14)
   14: END (0)

What the programmer almost certainly intended, but neglected to include 
a crucial colon, was silently turned into something completely different.

However, currently

$ perl -Dr -le 'qr/(?[ [z[:ascii]])/'
Compiling REx "(?[ [z[:ascii]])"
Unmatched ':' in POSIX class in regex; marked by <-- HERE in m/(?[ 
[z[:ascii <-- HERE ]])/ at -e line 1.

When I was designing the (?[...]) extension, I did not feel so 
constrained by current practice, and put things like this in.  Some 
users have requested that all [...] behave like this, and this addition 
meets that request, via a selectable feature.

I propose that we document this feature as being extensible so that we 
may, from time to time, add new strictures to those currently 
implemented without fear of breaking backwards compatibility.  That is, 
we may make things stricter in future releases to catch more errors that 
we don't do now, and if your code no longer compiles in that release, 
you'll have to change it rather than we back out the stricter 
strictures.  I'd rather not, but if that means marking this as an 
experiment, so be it.

What the pragma would mean initially is yet to be fully determined.  It 
would mean at least all the things that (?[...]) prohibits or warngs 
about.  I would poll this list before adding things that I was not 
completely confident about.

• RFC: "use re 'strict'" by Karl Williamson
• Re: RFC: "use re 'strict'" by Ricardo Signes
• Re: RFC: "use re 'strict'" by Eric Brine
• Re: RFC: "use re 'strict'" by Ricardo Signes
• Re: RFC: "use re 'strict'" by Karl Williamson