> We've been wondering over here at sensors what the point was of unique > IDs. nobody seemed to use them. I think we have to make a difference between adapters and chips here. It looks quite obvious that adapters' IDs may have a use (as this is the case now). What I think Greg was suggesting is that we could drop IDs of _chip_ drivers, because those will hardly ever get used. Although, I don't think was thinking of deleting the id field in the structure definition, only not to fill them when we don't need to. As time passes, I tend to agree more and more. I just have some difficulties submitting a driver with no ID, because they all have one in both 2.4 and 2.6 so it feels strange to change now. But I admit it isn't rational. I think it would be acceptable to stop filling the id field by default, and only fill them when we need them to be filled. > Your application is an obvious example of their purpose. > On the Voodoo, the DDC bus should have a separate ID from the I2C bus. > Then you could distinguish them without a strcmp. Correct. > Perhaps we should add IDs to 2.6? And have 2 different IDs in the 2.4 driver as well, for we want to be consistent. -- Jean Delvare http://www.ensicaen.ismra.fr/~delvare/
