While I applaud stevieb's approach and might even consider it a best practice, holding oneself back to versions of Perl that were EOL'd a decade or more ago is akin to saying new versions of Perl do not matter. Moving from 5.6 to 5.26 is trivial compared to, say, moving from python 2 to 3. And if we're in a marketshare war, I argue expending effort to support old code is philanthropic but invisible.

I think //= is, at this point, reasonable to expect. That means 5.10. I know there is a lot of 5.8 out there, I'm unhappily stuck on it at work but not for much longer. I don't personally feel the slightest responsibility on the cusp of 2018 to support it.