Apple Acquires Primephonic

[Spotted Horses]

To summarize my own method, I use iTunes (not Apple Music) for my pop music, but not for classical. When I started ripping I was using foobar2000 on a PC, and the tags it gets from the free databases often utterly useless. Now I use XLR on MacOS, which outputs FLAC files with cue sheets and has access to the real gracenote database, which not nearly as bad as the free databases, but still not acceptable for classical music. If I am ripping to Mahler Symphony No 5, for example, "Mahler" can show up in the track title, in the artist field, in the composer field. The artist field can be the composer, the conductor, etc. Mostly a hodgepodge of information is found in the track name.

So I made my own tagging system which is very coarse grained - no effort to tag individual tracks (I just accept what gracenote gives me). Each CD or CD set is in its own folder named for the record label and selection number. So for this release:



All of the audio files, as well as a pdf file containing a scan of the booklet, is in a folder named DGG423696-2. (Having the folder named this way means that if I find a CD on another site I can quickly check to see if I have it ripped by looking for the folder name.) The folder also contains a file called contents.txt, containing my own tags. The contents.txt file for this release looks like this.

label : "Deutsche Grammophon"
catalog : "423696-2"
acquisition_year : "2000"
n_discs : "1"

composer_first : "Igor"
composer_last : "Stravinsky"
conductor : "Paul Sacher"
ensemble : "Philharmonia Orchestra"
ensemble_type : "Violin and Orchestra"
recording_year : "1988"
performers : "Anne-Sophie Mutter:Violin"
works : "Violin Concerto::Violin Concerto"
program : "reset"

composer_first : "Witold"
composer_last : "Lutoslawski"
conductor : "Witold Lutoslawski"
ensemble : "BBC Symphony Orchestra"
ensemble_type : "Violin and Orchestra"
recording_year : "1988"
performers : "Anne-Sophie Mutter:Violin"
works : "Partita::Suite"
works : "Chain 2::Suite"
program : "reset"

At the very top are the tags for the "release" (defined as a CD or CD set as released by the label). The annoying thing is a release can have recordings by different composers, different conductors, etc. So each release contains a list of "programs." This release has two programs, one with Mutter performing the Stravinsky violin concerto, the second with Mutter performing two works by Lutoslawski.

Now this content.txt file would be pretty useless, except that I have written a c program that scans the directory tree for contents.txt files and compiles a list of releases, each of which of which contains a list of programs, each of which contains a list of works (and the other data, conductor, performer, ensemble, etc). When my program is finished compiling the list it produces a series of HTML files in which the programs have been sorted by composer, conductor, performer, ensemble, genre. So I can see a list of all programs featuring Mahler, all programs featuring George Szell, all programs featuring the Chicago Symphony, all programs featuring viola concerti, all programs featuring cello and orchestra, etc.

The top of my list of recordings feature Anne-Sophie Mutter is attached. If I click any of the album cover images it opens a detail page for the individual release.

It works well for me, and sometimes I do computer programming as a sort of a hobby so I enjoyed the process.

[Pohjolas Daughter]

You are definitely right in your first statement: it does require a lot of manual input.
But spread out over 20 years, it's not so bad!

I only supply 8 tags, too, and 6 of them are 'recording-wide' (that is ARTIST, COMPOSER, ALBUM, GENRE, COMMENT and YEAR or DATE apply to all tracks of a particular recording). Only 2 are track-specific (i,e., TRACKNO and TITLE). So, for a lot of CDs, you really only have to type in the track titles to any great extent. So it's manageable. But yeah, I wouldn't want to have to tackle several thousand CDs ab initio these days 

So what kind of things do you enter into your comment section?  And how picky do you get re the artist input?  You could go bananas (and I did so to speak some years ago) entering in the various singers and vocal parts for say Bach cantatas for instance. 

To summarize my own method, I use iTunes (not Apple Music) for my pop music, but not for classical. When I started ripping I was using foobar2000 on a PC, and the tags it gets from the free databases often utterly useless. Now I use XLR on MacOS, which outputs FLAC files with cue sheets and has access to the real gracenote database, which not nearly as bad as the free databases, but still not acceptable for classical music. If I am ripping to Mahler Symphony No 5, for example, "Mahler" can show up in the track title, in the artist field, in the composer field. The artist field can be the composer, the conductor, etc. Mostly a hodgepodge of information is found in the track name.

So I made my own tagging system which is very coarse grained - no effort to tag individual tracks (I just accept what gracenote gives me). Each CD or CD set is in its own folder named for the record label and selection number. So for this release:



All of the audio files, as well as a pdf file containing a scan of the booklet, is in a folder named DGG423696-2. (Having the folder named this way means that if I find a CD on another site I can quickly check to see if I have it ripped by looking for the folder name.) The folder also contains a file called contents.txt, containing my own tags. The contents.txt file for this release looks like this.

At the very top are the tags for the "release" (defined as a CD or CD set as released by the label). The annoying thing is a release can have recordings by different composers, different conductors, etc. So each release contains a list of "programs." This release has two programs, one with Mutter performing the Stravinsky violin concerto, the second with Mutter performing two works by Lutoslawski.

Now this content.txt file would be pretty useless, except that I have written a c program that scans the directory tree for contents.txt files and compiles a list of releases, each of which of which contains a list of programs, each of which contains a list of works (and the other data, conductor, performer, ensemble, etc). When my program is finished compiling the list it produces a series of HTML files in which the programs have been sorted by composer, conductor, performer, ensemble, genre. So I can see a list of all programs featuring Mahler, all programs featuring George Szell, all programs featuring the Chicago Symphony, all programs featuring viola concerti, all programs featuring cello and orchestra, etc.

The top of my list of recordings feature Anne-Sophie Mutter is attached. If I click any of the album cover images it opens a detail page for the individual release.

It works well for me, and sometimes I do computer programming as a sort of a hobby so I enjoyed the process.

Sounds pretty cool (from my totally inexpert point of view anyway)!    ).

PD

[lordlance]

Totally agree with you there (except I'd replace 'often utterly useless' with 'always utterly useless'  )

I part with you at this point, because I think it important that the logical data and the physical storage follow identical formats. If I see 'Symphony No. 4' being played in my player, I want to be able to see a folder called 'Symphony No. 4' in my file manager on my PC. It's a discoverability issue for me. Most times, I play music via a player that is using metadata to organise the music; but sometimes I just think to myself 'I want to play a particular recording', and the quickest way to do that is maybe to use a file manager to locate the relevant file and just right-click it and say 'play'. I need the file manager to be able to navigate the same way the music player organises its 'library': so for me, the metadata prescribes the physical storage, and vice versa.

I haven't read your entire article on tagging so you'll have to forgive me if it's covered but based on what I read in Axiom #3, if you want to sort tracks by artist, you would be reliant on the comment tag? The issue there which I see (and which seems a bit fundamentally unsolvable) is that you can't sort tracks by conductor/soloist if you so desired. Bohm and Pollini made a few recordings together for example but not necessarily with the same orchestra so even within a consistent naming pattern, they would be quite far apart.

[lordlance]

That's the same three recordings of Britten plus Baker that I was displaying before -but now the principal artist is determining the sort/display order. Now, I only put the conductor in my PERFORMER tags, and I think it would get really messy if I allowed combos of conductors and performers in the same field (because 'Bohm/Pollini' would appear toward the top of the list and 'Pollini/Bohm' would be lodged somewhere in the middle. But the principle is that if your music player lets you be flexible in this sort of way, then you can get imaginitive with what data you insert where, without completely violating any of my alleged 'axioms'

At the end of the day, your music collection/database needs to answer queries you make of it. So if, for you, the conductor and pianist (or chief soloists) are the common way you find music to play, then yes: you'd probably want to promote those into readily  displayable fields that common music players tend to default to using. For me, composer is king, prince and court... I'd choose to play Beethoven or Britten, and I couldn't (usually) care less who's performing it. Therefore, for me, composer gets top-billing and the lesser details get free-form text comment treatment (though the 'principal artist' usually gets a mention in the album tag, by surname, because that's important to distinguish between different recordings of the same piece by the same composer at different times). It seemed to me when I wrote those axioms that this would be the general approach of most classical music listeners, which is why I wrote as I did. But there will always be exceptions.

And therein lies the major problem: For me classical music has always been heard virtually every time through the prism of the artist even when I had just started and had no idea what the differences between conductors could be. Before trying to implement it, the issue has to be solved logically first. There are times when the artist is important and times when the soloist is important (for example when listening to performances by Idil Biret the conductors can often be random people which I have never heard of.)

Then of course the issue of arrangement - Bohm/Pollini/VPO|BPO throws a wrench in the naming scheme of Bohm/VPO and Bohm/BPO. I do not have answers to these questions...

But since I constantly delete and download new music I have much less utility out of perfecting my library. I stick to folder and file names to help me out...

I have stuck to Soloist/Conductor/Orchestra/Year (for live performances) and that works. The issue is when downloading new files and you end up with file names like 'Masur LGO Beethoven Third' or 'Estrada Beethoven Sinfonie Nr. 3" where you can only sigh and hope that you catch such bad file names when you download them and rename them there and then.

I use foobar2000 but the search results are still often messy because it tries to match your search query to every single metadata container and that's how you end up with things like this:



It's not particularly good but I hate VLC and have tried musicbee. I don't know what else to try...

Also which music player do you use?

[Spotted Horses]

I personally couldn't live with data being external to the thing it's describing like this. I don't even have album art as a standalone 'folder.jpg' or similar. If it's data of any sort, it's going to get written into the FLAC file itself. That way, it can't be deleted or corrupted or otherwise rendered 'wrong' without the music file itself similarly being damaged. Is there a reason you don't just create a bunch of custom tags to store this stuff within the file itself?

Actually I like having the data in distinct files. When I run my indexing program it makes a tree of HTML and related files (album covers, pdf booklets, cue sheets) that I copy to other computers (which don't have access to the media files) so I can browse my collection anywhere.

[Spotted Horses]

Fair enough. I used to have album art separate from the files too, (you know, the Windows 'folder.jpg' type of thing), until one day a 'helpful' music player decided to obliterate the lot by over-writing them with freshly fetched art from the Intertubes (all of which was wrong, naturally!). I've also done a 'move *.flac \\somewhere' and entirely forgotten the existence of the related JPGs. Thus, these days, I embed -and my software simply reads the tags with metaflac and constructs whatever it needs from there. If I was creating an HTML catlogue, including art and booklets and so on, I could still do that with embedded data. (For example, metaflac --export-picture-to=//somwhere//something.jpg). But I just won't risk having different bits of related data walking off in different directions at some point: call me paranoid, I guess!

For the 'browse from anywhere' functionality (though including the ability to actually play my music via the Internet), I use Emby. Kept me happy when last in Scotland, anyway! Also handy for remotely watching movies and even browsing photo albums when visiting family!

My preferences are based on a universal distrust of media players. When I decide to listen to something I copy the media files to a cache folder and load it into my media player library. When I have finished listening I delete from the media player library and from the cache folder. At any given moment, my media player thinks I only have a dozen or so CDs. It works for me, I wouldn't assume it would be satisfactory to anyone else.

[Spotted Horses]

Just to clarify, my main catalogue is in a SQL database, populated by a lot of 'find *.flac' and 'metaflac --export-to' commands: my primary media player is merely a shell script that interrogates the database, then hares off to the file system to play whatever it finds there. So I get your point about being cautious around media players generally.

Also: Emby is not really a media player as such. It's basically a MySQL database with an HTML front-end and some dynamic transcoding capabilities that again otherwise merely reads from a hard disk folder.

PS. I meant to asK: why do you use media players that way? What is it about them you distrust?

If I had a working knowledge of proper database software I'd probably have used that.

I prefer not to be dependent on media players because in the past I have not found their search functions to be useful. I found it more satisfying to develop my own tagging system, rather than try to reverse engineer the tag handling of a media player. Just my preference, I wouldn't claim it makes sense to anyone else.

[Karl Henning]

Couldn't help misreading the acquisition as Pimp-phonic.

[Spotted Horses]

I understand. It's just that when you said you 'distrust' them, it made it sound as if you'd had bad -i.e., damaging, music-loss type- experiences with them, which had me wondering!

I had the experience of Winamp crashing my Windows computer.

I've been wanting to get a working knowledge in Python, the thought occurred to me to re-implement my indexing program in Python instead of c. But the time is not there.

[staxomega]

Currently I have a large collection of classical music (CDs, mostly ripped, and lossless Digital downloads). I effectively have my own private streaming service that I don't have to pay for. The cost of purchasing the occasional item not in my library amounts to less than the monthly cost of a streaming service, so it doesn't make economic sense to me, even putting aside not liking the idea that I am supporting tech companies rather than creators. If my collection were destroyed in some sort of natural disaster I guess I'd have to re-evaluate the economics. But the fact that your favorite content can disappear from the streaming services without warning is close to being a deal-breaker for me.

Someone did this with a large classical collection via Roon. $120/year is a bit too steep for me for just accessing my local collection, and I like JRiver's interface. I would like to do what you propose, but instead of having all the music accessed by the server (local streaming) have some sort of synced backup type system with that server among every room that has a stereo system, so essentially each computer has access to my full music collection. I don't think JRiver's BubbleUPnP or whatever is supposed to be as good as RoonCore.

[lordlance]

Well, my answer to those questions is: filter and search. If I was really determined one day that I wanted to listen to Böhm with Pollini, though that's not something I would routinely do, I can nevertheless still do this:



So it turns out I do have three recordings with that combo! And you'll note it took me 34 milliseconds to find them. To play a specific one, I can then just issue this command to my preferred music player:

Well I think I need to search by file name OR title because either could be where the file name is. I haven't learned foobar syntax yet. I suppose I can try to mess around with it.