An efficient alternative to:

while (<DATA>) {
    $_ .= <DATA>;
    s/^0\s*(.*)\n\s+(.*\n)/$1 $2/s;
    print;
}
[download]

is not to even loop at all.

{  local $/ = undef;
   $_ = <DATA>;
    s/^0\s*(.*)\n\s+(.*\n)/$1 $2/s;
    print;
}
[download]

caveat is that it eats up memory if you have an extremely large file.

Cheers,
Kristina Update: See BrowserUK's post immediately below. /s would allow the "." to match the newline which I don't want to happen here, but /m will treat the single string as multiple lines and /g will let me match more than once before quitting.