This is definitely getting into TIMTOWTDI territory. If you suck in the entire input using a slurp (see $/), then your existing regular expression would work.

Rather than using the character class [\s\S], I think intent would be more obvious using . combined with the s modifier: if ($_ =~ /start(.+?)end/si)

I would probably also add word boundaries around your start and end, so you don't get false positives from words like 'send': if ($_ =~ /\bstart\b(.+?)\bend\b/si) Even better, if you know that start and end are on lines by themselves, use the m modifier to say as much: if ($_ =~ /^start\s*\n(.+?)^end\s*$/smi) Regular expressions play best when they are tightly constrained.