Threaded Mode | Linear Mode

DavidHB · (This post was last modified: 30-04-2014 13:56 by DavidHB.)

With MinimServer update 25 in mind (which is such an important release it is surprising that the version number was not advanced to 0.8.2), I have been re-reading the interesting discussion on the various methods of describing compositions and artists in metadata tags. Many thanks to all the contributors who provided thoughtful and informative contributions which IMO show the forum at its best. And thanks also to Simon, because without MinimServer much of the discussion would be academic.

Throughout the thread, it is a given that there is no one 'right' way to tag classical music. From the posts it also appears that most contributors regard their tagging system as work in progress, as I certainly do. The issue then arises as to how one selects between the various options, to suit one's own purposes and preferences. On this, I offer the following thoughts, in the hope that others might have useful contributions to make.

Firstly, I think we have to be realistic about what tagging and the associated indexing on the server can and cannot achieve. Sadly, there is no easy way to have digital music provide the amount of associated information we are used to getting on LP covers and in CD booklets. (The Bubble control points have an interesting means of associating a scanned .pdf file with a track, but the prospect of scanning hundreds or thousands of CD booklets is not one that most people will happily contemplate.) The metadata we store can, in practice, be used only for

browsing or searching for tracks to add to the playlist;
displaying details of the track on the 'now playing' screen (or equivalent) of the control point.

Unless we have realistic hopes that the features of our preferred control point will improve, there is little point in adding data that the control point will not use. With the capabilities MinimServer already provides, it is control points that will be the limiting factor for the foreseeable future; indeed one major improvement to a number of control points I have tried would be just to allow the display of the MinimServer browse tree as is, and without any caching.

Another factor which will affect our choice of tagging scheme is our own willingness and ability to impose order and consistency on the data. With classical music, this is quite a lot of work. Few of us would be without the online metadata libraries used by ripping programs such as dBpoweramp, but the fact remains that the downloaded data almost always requires a significant amount of editing by the user. For instance, if you want to browse your library by composer as I do, each composer's name has to be spelt consistently throughout the library - you have to decide whether you are going to use, say, Sergei Rachmaninov, Sergey Rachmaninov, Serge Rachmaninoff or Sergey Rakhmaninov (all of which are in use), and then stick to your choice. Thankfully, MP3Tag provides an easy means of applying a tag value to a whole selection (possibly hundreds) of tracks at once.

A third issue is the way in which different tags relate to each other. As metadata tags are not defined hierarchically or relationally, this may require information to be duplicated in several different tags in some instances but not in others. For instance, in my own scheme, I have do not include the composer name in album and track titles, but I typically do include the name of the work in the track title and in the Composition and Group tags. Decisions of this kind require assumptions to be made about how the data set as a whole is going to be browsed (or searched) and viewed. This is why the advice given earlier in this thread, that new users (or those reorganising a collection) should start by experimenting with a small number of albums, is so sensible. MinimServer Update 25 adds a whole host of new possibilities for creating these relationships between tags and for displaying tags in different ways.

I suppose that the key point I am making here, and which has at least been hinted at in previous discussion, is that the way we populate individual tags is a product of assumptions on how the data to be used and how, therefore, the tags relate to each other. These assumptions are usually developed over time; the data set associated with classical music is complex, and only a few fortunate people will know from the outset what tagging scheme best suits their needs. My own tagging pattern has reached the stage at which I know how I wish deal with composer and composition data, but am still uncertain as to precisely what place artist data occupies in the general scheme. I guess I am not alone in having aspects of my scheme that are better defined than others.

I hope these thoughts are useful.

David