I would agree that if QoS were the only problem that needs to be solved it could be acceptable to carry QoS information in the LDP messages that are used for PW setup (albeit that it is somewhat ugly since it makes LDP look like CR-LDP which has been depreciated).
However, there are other issues that need to be solved. Several people on this mailing list suggested that we need to address PW stitching. It seems to me that stitching of PWs requires a signaling protocol that establishes a multi-hop PW through switches that switch based on the PW label. RSVP is in my view the only reasonable solution.
Hence, if we don't look at one problem at a time, but at the entire problem space, I think we need to adopt RSVP as a way to establish PWs. It would solve the problems of both QoS and stitching.
Actually PW stitching can be implemented entirely using existing pseudowire
signaling procedures (using LDP or L2TP) without any new protocol or extension.
The signaling protocol is not an issue.