$ perl -le '$_="1 2 3 4 5 6"; s/^((?:[^\s]*\s){3}[^\s]*)\s/$1\n/ and p
+rint'
1 2 3 4
5 6
[download]

That might be more readable as:

s/^([^\s]*(?:\s[^\s]*){3})\s/$1\n/
[download]

or expanded:

s/^([^\s]*\s[^\s]*\s[^\s]*\s[^\s]*)\s/$1\n/

The idea is to match anything that is not a space followed by a space and 0 or more non-spaces 3 times and capture it. Then match another space and replace all of that with what you captures and a newline at the end.

Expanding and commenting gives us:

s/
    ^             # Start at the beginning.
     (            # Start capturing.
       [^\s]*     # 0 or more non-space.
       \s [^\s]*  # A space and 0 or more non-spaces... once.
       \s [^\s]*  # twice.
       \s [^\s]*  # three times.
     )            # Stop capturing.
    \s            # Match another space.
/$1\n/x           # Replace with what we captured and a newline.
[download]

Of course, that all looks rather ugly with the negated char classes written like that. You should probably use \S instead. So, finally:

s/^(\S*(?:\s\S*){3})\s/$1/n/

-sauoq
"My two cents aren't worth a dime.";