Transformations currently has the capability to load text from files, but it could easily get that content from online sources as well. For example, this could be done with cURL like in parser_csv.inc in Alex Barth's CSV parser module. Another (less powerful) option would be drupal_http_request(), which returns the file contents in the $result->data field in case the request was successful.

(Of course, there's no reason not to support both methods.)