Defining which file formats a VRML implementation "must" support (where
"must" == we tell authors they can count on these formats...) is still
an open issue. It seems like JPEG and GIF are the popular HTML image
formats; do we need others? (by the way, does JPEG support alpha
channels in images?)