On Fri, 2011-11-11 at 12:42 +0100, David Henningsson wrote: > On 11/10/2011 09:34 PM, Tanu Kaskinen wrote: > > I got a bit confused about the path naming, or more specifically about > > the uniqueness of the path names. I didn't have time to double check, > > but it seemed like in the old system there was no reason to have the > > path_set_make_paths_unique() function at all, because it looked like the > > path names would be unique anyway (within one path set). I'm pretty sure > > that I've missed something... > > Two different path files can have the same name, as "name" refers to the > "Name" key in the "General" section, not the file name. Does that clear > things up? Yes, thank you. Having the "name" option in General seems pointless to me... Complexity without gain. If the "name" option wouldn't exist, there would be 1:1 mapping between the path names appearing in PA logs and the configuration file names, which would be quite nice. Also, there wouldn't be need to care about making the path names unique after probing (port descriptions would probably still have to be checked that they are unique). For those reasons I'd like to remove the "name" option. Do you have anything against that? > > The path set should be created already at configuration parsing time. > > There shouldn't be need to create anything at probe time - I think the > > purpose of probing is just to remove those paths that are not available. > > It is a slight optimisation to create it here, this way we only create > path sets for profiles that are supported. Ok. I'd still like to keep the clear phases of configuration parsing (structure creation) and probing (removing unavailable stuff). So what do you think? Your comment didn't really tell me whether you agree or not. -- Tanu
