Using split() with a limit is indeed better. But I still feel like elaborating some on your regex, for the heck of it. ;)

In my opinion, your regex is too restrictive. It requires a colon and something after it. This isn't the way split() works. To make it more analogous, I'd make

/([^:]*):?(.*)/s;    # Keeping it simple
/([^:]*)(?::(.*))?/; # Making it even more analogous.
[download]

(Note also the s modifier that's very easy to forget.)

Of course, it's not perfect. There are issues about the length of the returned list; if the "extra" trailing element (that split() would not even return) should be undef or ''; etc.

The second regex returns '' as second element if there was a colon, undef if there was no colon. This is as close as one gets I believe.

Just for fun:

grep defined, /([^:]*)(?::(.*))?/;
[download]

I can't find any differences between this and split(/:/, $_, 2), except for scalar context execution, were you get rid of the old @_ behaviour.