Separating releases based on relationships inferred from matrix codes

Recently, I put in a merge edit for two releases (this one: Edit #108784804 - MusicBrainz). It’s the kind of merge that usually raises no doubts: same date, tracks, release country, cat. no., barcode, and disc ID.

Another editor voted I against it, saying they are different pressings because they have different matrix codes. It seems he’s getting these codes from Discogs, although there is no evidence for them (no images, just lists of unverifiable codes). It checked the forum and it seems most people agree that’s not enough to separate releases, and mentioned that. He then added different pressing relationships, based (as far as I can understand) on these codes. There are two things that bother me here:

  1. Copying and pasting long lists of codes from Discogs, without knowing if they’re accurate and with no way to verify them (no CDs to check, and no images on Discogs).
  2. Inferring relationships from these unverified codes and editing based on that. Which makes two identical CDs separate releases.

Is the general view is that this is fine? I have no problem cancelling the merge, I’m just confused about this.

3 Likes

Different pressing plants are generally different, separate Releases. Though I agree with the uncomfortable copy\paste of long lists of unverified matrix details from Discogs. Seems wrong to me and I like my matrix details. Discogs are often tweaking those messy long lists and removing errors. Blindly copying them seems a bit weird.

4 Likes

I don’t really have an issue with this. If somebody add a new release and says “hey, this looks identical to the other one, but it has a different matrix code, which proves it’s from a different pressing plant; here’s the image of the matrix code, and here’s the disc ID I got from this CD”, I wouldn’t challenge that.

But when you just blindly copying-pasting from Discogs, how do you know that’s the case? And how do you know which disc ID corresponds to which matrix code/pressing?

1 Like

I do basically agree with you, and I am a matrix addict.

Looking at those two releases, the one with three discIDs listed is the bigger mess of confusion. All of those are pre-NGS and therefore none attached to that actual release. They should all be removed really. NOTHING pins any of those three discIDs to that actual release.

The release with a single discID looks a bit more trustable. That was added in 2018. The discID and discogs link came from the same person so I would totally trust those being paid up correctly. (Scroll back to the first page of the edit history for that one)

Personally when I add details from my own CDs in hand, I type in my Matrix details along with the matching discID as a matching set in the annotation.

Edit: I don’t know how I do typos like that… “paid up correctly”? Makes zero sense, but spelt correctly. “added correctly” is what I meant…

6 Likes

And that editor seems to be active still.
@blackteadarkmatter Did you try to contact @SanMarc.60?

Copying my comment from the edit note discussion here, because it deserves repeating:

I fully agree with you that copying matrix codes from associated Discogs links is too unreliable to base editing decisions on. Ideally, that information is documented in the uploaded cover art and the annotation field. This why it is super useful to list the DiscID together with the matrix information in the annotation, like is done in this release:

In addition, it is good practice to support that information with scans from the matrix/runout information, as I have done for example here.

@UltimateRiff wrote a helpful guide for assisting people in understanding and documenting this information.

spread the word :smiley:

3 Likes

Also restating from the edit notes:

I’m cancelling this edit since it seems clear to me the consensus is that there is a good quality release for a specific pressing, and it shouldn’t be merged with the remaining ones.

I actually can agree with this in this specific case because somebody pointed out the same person worked on both sites, so it’s likely that the disc ID does match the pressing/matrix on Discogs. (I’m not a Discogs editor, so I don’t know how to see who created the entry, and I can’t see the history.)

But, if anything, I’m more convinced we shouldn’t do this, not based on Discogs matrix code lists. On [MusicBrainz] we have disc IDs, and that’s our main means of identifying releases. We have no way to know if a specific pressing in Discogs actually corresponds to the disc ID on the MB entry it’s linked to, because many people just import Discogs releases to MB, 1:1, without actually having them or checking if they’re different. That is, you may be carefully annotating the release with pressing information and matrix codes, but the disc ID in attached isn’t even related to the pressing.

I don’t have a problem with somebody who has a specific pressing adding it as a separate release, adding the pressing relationship and the matrix codes from their disc – if they care about it, and many people do. But simply copying them from Discogs with no way to check if it actually corresponds to the disc ID still feels like a bad idea to me.

(…)

While I personally don’t care for matrix codes, it would be very helpful if we had a field for matrix codes at the release level, with a link to a wiki page explaining how to look up information based on those codes. Encouraging people to add matrix codes to releases when they submit disc IDs would help undo much of this confusion.

(https://musicbrainz.org/edit/108784804)

1 Like

You should register. You don’t have to make any edits, but you can then see the history.

If the correct version is linked, the information is reliable. I treat a MB release as if it were this version. I usually find too little information on MB to clearly identify it. I rarely add releases that I don’t own, but when I do, the new MB release becomes exactly that release. And I add all relevant features on MB so that other editors can add a disc ID for their matching release.

:+1:

2 Likes

Medium level: https://tickets.metabrainz.org/browse/STYLE-781

2 Likes

You are correct in doing this, but not everyone does the same. When I look into the edit history of messy release groups, it’s often clear many people are just trying to import Discogs releases into MB, often without additional notes or any way to confirm info from Discogs. Then somebody who actually owns the CD attaches it to a release, based on the basic info on MB. What you do is what we should do, but we can’t assume everyone does that, we can’t assume the linked Discogs page is for the same release, especially if we separate by pressing, which most people won’t even notice. MB is its own thing, we can’t treat the linked DC page as additional information for an MB release. A link just means an MB editor thought that was the same release. To use any information, you need to verify it, by comparing the scans (if you can make out anything) and possibly matrix code.

I also add Discogs links to releases, and also look up information on Discogs. It’s just a massive resource, and very useful, but we can’t just assume all links are accurate, and all information correct.

3 Likes

And when I work on it, it becomes this exact release. I can’t verify disc IDs (if it’s not mine), but I look on other information, if there is something, and if pictures are attached, I try to find the correct Discogs release, if there’s also a disc ID I ask the editor who submitted it if their release matches the images, and so on…
If there’s a disc ID (attached by a different editor) and no further information, only the link, I also try to get answers, but I would make it the linked version and would risk having a false disc ID. Most often, there are already a lot of them anyway.

Exactly! But not “possibly” matrix codes. Matrix codes are the key part of my research. The code (written by LBR, not the “stamped” codes from duplication) has exactly one corresponding disc ID, whatever is printed on the disc or the cover.

Never. All we can do is try our best to sort out the mess. That’s sometimes tiresome… :slight_smile:

5 Likes

While I started with working over the disc IDs I have added (mostly to own submissions) by adding matrix/runout and SID data to the notes of these adds I realized this: Edit #101204638 and Edit #101128940 have identical disc ID, but completely different data, release groups, manufacturers and dates.
My question: How strong can we rely on disc IDs to separate releases?
… and what is LBR?

1 Like

It is pretty rare to get two CDs of 15 tracks to have all those tracks of an identical length to produce an identical DiscID but different content. Rare but not impossible.

I’d put it at tiny fractions of a percentage chance. I have something like 800 CD (separate discs) I’ve added to MB and I think I only seen it happen once. And that was a single with only a couple of tracks.

Am I missing something? That is the same track list in both cases, just in different boxsets. This makes sense they are the same discID.

Note that Sony DADC and Avarto are basically the same factory, just different owners at different times. That looks to me like Universal just change the content of the box at some stage, but still used the same source for CDs.

A DiscID can last for decades and not change. Not unusual for it to outlive the ownership of the factory.

3 Likes

In fact, in the above post, I tried to choose the words exactly:

The code (written by LBR, not the “stamped” codes from duplication) has exactly one corresponding disc ID

That’s not true in the other direction. New glass master versions from the same plant reuse the TOC from earlier versions in most cases. Of course, it is also possible to transfer this information to another plant. But this was rarely done. It doesn’t seem to be too difficult to set the index points anew.

However, in your case it looks like this was done from Arvato in Gütersloh to Sony in Thalgau.

This shows how valuable this information is. All CDs with this matrix will necessarily generate this disc ID.

1 Like

It means laser beam recorder. :sunglasses:

2 Likes

I think in this context it would be rather interesting to have the matrix data from https://musicbrainz.org/release/1497341a-4191-44c3-a563-32181b80dde3 too.
I only have copies from the ones quoted in the above edits.

1 Like

Sadly there’s no matrix given on Discogs, so we can only assume but we will not know for sure.

1 Like