Usually, you don't add records to your full-text index for documents that are not text: images, audio, binaries, etc.
However, some users wish to have records in the full-text index for all documents, regardless of their type. This way, any document can be found by searching for the URL, or a part thereof.
This is also needed to support WERA-like functionality where the Nutch(WAX) index is used as the Wayback index as well. In this case, all documents need a record in the full-text index.