According to the perlnumber I'm reading from my 5.8.5 documentation, +, along with many other binary arithmetic operators, "will attempt to convert arguments to integers" - if that doesn't keep precision, then "arguments are converted to floating point format and the floating point result is used." Since conversions are covered above, we go back to see how "decimal string" is converted to "native" types (floating/integer).

Is that what you're looking for, or am I misreading your question, the documentation, or both?