Earlier quoted context omitted.
Well, only if someone tries to go to http://XXX.XXX.XXX.42 , which seems rare. I can't think of a way to tell whether XXX.XXX.XXX.42 is an IP or a website from that string alone. You could require that no subdomain has four "parts" (i.e., A.B.C.42), but that's harsh. You could require that somewhere in the URL for a four-part numeric-TLD URL there be a character. You could also require that at least one of the parts…
reading the rules from an rfc mentioned on the wiki, if XXX.XXX.XXX.42 is a valid IP address, the client should try to use that first, if not it should only then try to use DNS to resolve.
But note that the RFC assumes there will never be a numeric TLD. Furthermore, it suggests checking the string syntactically to determine if it's in dotted decimal form, remaining somewhat ambiguous about what to do if the hostname looks like IPv4, but the connection fails (it seems unwise to do a DNS lookup for every IPv4 address that's offline). Since IPv4 is relatively easy to validate, it would have made a lot more sense for a numeric TLD to select a number outside the range of 0-255, such as 4200. Then it's obvious that 1.2.3.4200 is a hostname (or harmless typo), and not an IP address.