RFC: Improve Guidelines on Merging Recordings

classic Classic list List threaded Threaded
120 messages Options
1234 ... 6
Reply | Threaded
Open this post in threaded view
|

RFC: Improve Guidelines on Merging Recordings

LordSputnik
Expected Expiration Date for RFC: Wednesday, 7th November 2012


This proposal is designed to improve the guidelines on merging recordings, at https://wiki.musicbrainz.org/Style/Recording.

It is felt that the current guidelines are too vague, and that they don't cover all cases where recordings might be merged. This proposal seeks to expand these guidelines and make them clearer.

JIRA Ticket (includes previous IRC discussion): http://tickets.musicbrainz.org/browse/STYLE-159
Wiki Page: https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines


This is my first Style Proposal, so if I've done part of it incorrectly or I've missed some part of the procedure, please let me know! However, I think everything is as it should be.

Kind Regards,
Ben (LordSputnik)

_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Calvin Walton-2
On Wed, 2012-10-31 at 17:29 +0000, Ben Ockmore wrote:

> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging recordings,
> at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they don't
> cover all cases where recordings might be merged. This proposal seeks to
> expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>
>
> This is my first Style Proposal, so if I've done part of it incorrectly or
> I've missed some part of the procedure, please let me know! However, I
> think everything is as it should be.

There's one minor issue with this in my opinion, regarding remasters: In
many cases, a release labelled "remastered" doesn't actually have any
audible difference from the original track other than a change to the
amount of gain - and that sort of minor gain adjustment is common in
things that aren't labelled remastered as well. IMO, that's not enough
change to be worth a new recording.

I think new recordings for remasters should be limited to particularly
"notable" remasters, or ones with noticeable audible differences.

One other thing that I think is worth mentioning: Recordings with
different numbers of channels (stereo/mono or stereo/5.1, etc.) should
never be merged.

--
Calvin Walton <[hidden email]>


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Kuno Woudt
In reply to this post by LordSputnik
On 10/31/2012 06:29 PM, Ben Ockmore wrote:

> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging
> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they
> don't cover all cases where recordings might be merged. This proposal
> seeks to expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines

Quoted from the proposal page:

 > The exception to this case is for changes that only affect the length
 > of the fade at the beginning or end of a track, where a new recording
 > should only be used when the difference in length is greater than 10
 > seconds.

I disagree with this.

If there is any audible difference between two recordings, one cannot be
used as a substitute for the other, and they should have seperate
musicbrainz identifiers.

e.g. if this fade in/out difference is the difference between the
version of a track released on a compilation vs the version on an album,
it would sound terrible to hear the one intended for one of those
releases in the context of the other release.  Any audible difference
should be a new MBID, otherwise the identifiers are
useless.

-- kuno / warp.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

onasis
See my recommendation about different edits of the same recording that is still to be accepted.
http://musicbrainz.1054305.n4.nabble.com/STYLE-151-more-Relationships-between-recordings-tt4640868.html
http://tickets.musicbrainz.org/browse/STYLE-151

2012/10/31 Kuno Woudt <[hidden email]>
On 10/31/2012 06:29 PM, Ben Ockmore wrote:
> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging
> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they
> don't cover all cases where recordings might be merged. This proposal
> seeks to expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines

Quoted from the proposal page:

 > The exception to this case is for changes that only affect the length
 > of the fade at the beginning or end of a track, where a new recording
 > should only be used when the difference in length is greater than 10
 > seconds.

I disagree with this.

If there is any audible difference between two recordings, one cannot be
used as a substitute for the other, and they should have seperate
musicbrainz identifiers.

e.g. if this fade in/out difference is the difference between the
version of a track released on a compilation vs the version on an album,
it would sound terrible to hear the one intended for one of those
releases in the context of the other release.  Any audible difference
should be a new MBID, otherwise the identifiers are
useless.

-- kuno / warp.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
In reply to this post by Calvin Walton-2
I do agree with you about remasters. I did actually think about including something like that earlier today, but wasn't sure if it was common enough to consider. Although, if the gain was increased too much, then the audio would be clipped, and audibly different.

Perhaps the wording could be "sonically remastered" to put the emphasis on a change in audio? Or "significantly remastered"? Though both of them sound a little clunky. A better wording would be much appreciated!

I would say that the cases 5 and 7 cover different numbers of channels - it's a different mastering and the audio has to be edited to make it suitable for different channel configurations.

On 31 October 2012 18:02, Calvin Walton <[hidden email]> wrote:
On Wed, 2012-10-31 at 17:29 +0000, Ben Ockmore wrote:
> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging recordings,
> at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they don't
> cover all cases where recordings might be merged. This proposal seeks to
> expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>
>
> This is my first Style Proposal, so if I've done part of it incorrectly or
> I've missed some part of the procedure, please let me know! However, I
> think everything is as it should be.

There's one minor issue with this in my opinion, regarding remasters: In
many cases, a release labelled "remastered" doesn't actually have any
audible difference from the original track other than a change to the
amount of gain - and that sort of minor gain adjustment is common in
things that aren't labelled remastered as well. IMO, that's not enough
change to be worth a new recording.

I think new recordings for remasters should be limited to particularly
"notable" remasters, or ones with noticeable audible differences.

One other thing that I think is worth mentioning: Recordings with
different numbers of channels (stereo/mono or stereo/5.1, etc.) should
never be merged.

--
Calvin Walton <[hidden email]>


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
In reply to this post by Kuno Woudt
That exception is intended only for changes in the fade, when there are absolutely no other changes to the audio. I added it in as it's the recommended threshold for assigning a new ISRC, from the ISRC handbook.

Can you explain a little more what you mean, or suggest an alternative wording?

On 31 October 2012 18:17, Kuno Woudt <[hidden email]> wrote:
On 10/31/2012 06:29 PM, Ben Ockmore wrote:
> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging
> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they
> don't cover all cases where recordings might be merged. This proposal
> seeks to expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines

Quoted from the proposal page:

 > The exception to this case is for changes that only affect the length
 > of the fade at the beginning or end of a track, where a new recording
 > should only be used when the difference in length is greater than 10
 > seconds.

I disagree with this.

If there is any audible difference between two recordings, one cannot be
used as a substitute for the other, and they should have seperate
musicbrainz identifiers.

e.g. if this fade in/out difference is the difference between the
version of a track released on a compilation vs the version on an album,
it would sound terrible to hear the one intended for one of those
releases in the context of the other release.  Any audible difference
should be a new MBID, otherwise the identifiers are
useless.

-- kuno / warp.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Frederik "Freso" S. Olesen
Den 31-10-2012 19:37, Ben Ockmore skrev:
>That exception is intended only for changes in the fade, when there are
>absolutely no other changes to the audio. I added it in as it's the
>recommended threshold for assigning a new ISRC, from the ISRC handbook.
>
>Can you explain a little more what you mean, or suggest an alternative
>wording?

I have a case;
https://musicbrainz.org/recording/b2d89a07-2efc-4e21-b737-00ffeefbf761
https://musicbrainz.org/recording/f5420ccf-4afe-42b6-9d2c-bf4b850c8053

These are both Backstreet Boys' Christmas Time on two different
compilations. They both have the same ISRC. However, they don't share
the same AcoustID. Notice the time difference - 4:08 vs. 4:16.

Both of the two start the fade-out at 4:00, and although one of them (I
can't remember which) is "louder", the kicker is that the :08 repeats
the refrain once during the fade-out while the :16 repeats it twice.

I disagree with these two recordings being merged, as they're definitely
not the same piece of audio. The AcoustIDs will also testify to that.
Even if they came from the same "master" from the studio. Which I think
is also what Kuno is saying.

--
Namasté,
Frederik "Freso" S. Olesen <http://freso.dk/>
MB:   https://musicbrainz.org/user/Freso
Wiki: https://wiki.musicbrainz.org/User:Freso

_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

onasis
In reply to this post by LordSputnik
It depends what you understand as Remaster.

http://en.wikipedia.org/wiki/Remaster

By definition Remaster is new master of the old one, so it is the new recording. So they can't be merged. Most of the remasters are digital versions of the analog, so this makes them totally different. The same about 2 different remasters of the same analog.

2012/10/31 Ben Ockmore <[hidden email]>
I do agree with you about remasters. I did actually think about including something like that earlier today, but wasn't sure if it was common enough to consider. Although, if the gain was increased too much, then the audio would be clipped, and audibly different.

Perhaps the wording could be "sonically remastered" to put the emphasis on a change in audio? Or "significantly remastered"? Though both of them sound a little clunky. A better wording would be much appreciated!

I would say that the cases 5 and 7 cover different numbers of channels - it's a different mastering and the audio has to be edited to make it suitable for different channel configurations.


On 31 October 2012 18:02, Calvin Walton <[hidden email]> wrote:
On Wed, 2012-10-31 at 17:29 +0000, Ben Ockmore wrote:
> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>
>
> This proposal is designed to improve the guidelines on merging recordings,
> at https://wiki.musicbrainz.org/Style/Recording.
>
> It is felt that the current guidelines are too vague, and that they don't
> cover all cases where recordings might be merged. This proposal seeks to
> expand these guidelines and make them clearer.
>
> JIRA Ticket (includes previous IRC discussion):
> http://tickets.musicbrainz.org/browse/STYLE-159
> Wiki Page:
> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>
>
> This is my first Style Proposal, so if I've done part of it incorrectly or
> I've missed some part of the procedure, please let me know! However, I
> think everything is as it should be.

There's one minor issue with this in my opinion, regarding remasters: In
many cases, a release labelled "remastered" doesn't actually have any
audible difference from the original track other than a change to the
amount of gain - and that sort of minor gain adjustment is common in
things that aren't labelled remastered as well. IMO, that's not enough
change to be worth a new recording.

I think new recordings for remasters should be limited to particularly
"notable" remasters, or ones with noticeable audible differences.

One other thing that I think is worth mentioning: Recordings with
different numbers of channels (stereo/mono or stereo/5.1, etc.) should
never be merged.

--
Calvin Walton <[hidden email]>


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
In reply to this post by Frederik "Freso" S. Olesen
And the proposal doesn't suggest that you merge them, since in that case the change in length isn't the only change - a refrain has also been edited out, so the exception doesn't apply.

On 31 October 2012 19:01, Frederik "Freso" S. Olesen <[hidden email]> wrote:
Den 31-10-2012 19:37, Ben Ockmore skrev:
>That exception is intended only for changes in the fade, when there are
>absolutely no other changes to the audio. I added it in as it's the
>recommended threshold for assigning a new ISRC, from the ISRC handbook.
>
>Can you explain a little more what you mean, or suggest an alternative
>wording?

I have a case;
https://musicbrainz.org/recording/b2d89a07-2efc-4e21-b737-00ffeefbf761
https://musicbrainz.org/recording/f5420ccf-4afe-42b6-9d2c-bf4b850c8053

These are both Backstreet Boys' Christmas Time on two different
compilations. They both have the same ISRC. However, they don't share
the same AcoustID. Notice the time difference - 4:08 vs. 4:16.

Both of the two start the fade-out at 4:00, and although one of them (I
can't remember which) is "louder", the kicker is that the :08 repeats
the refrain once during the fade-out while the :16 repeats it twice.

I disagree with these two recordings being merged, as they're definitely
not the same piece of audio. The AcoustIDs will also testify to that.
Even if they came from the same "master" from the studio. Which I think
is also what Kuno is saying.

--
Namasté,
Frederik "Freso" S. Olesen <http://freso.dk/>
MB:   https://musicbrainz.org/user/Freso
Wiki: https://wiki.musicbrainz.org/User:Freso

_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Sheamus Patt
In reply to this post by LordSputnik
On 10/31/2012 01:29 PM, Ben Ockmore wrote:
Expected Expiration Date for RFC: Wednesday, 7th November 2012


This proposal is designed to improve the guidelines on merging recordings, at https://wiki.musicbrainz.org/Style/Recording.

It is felt that the current guidelines are too vague, and that they don't cover all cases where recordings might be merged. This proposal seeks to expand these guidelines and make them clearer.

JIRA Ticket (includes previous IRC discussion): http://tickets.musicbrainz.org/browse/STYLE-159
Wiki Page: https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines


I think it's going too far to consider recordings different simply because one is analogue (e.g. Vinyl) and the other digital (e.g. CD). This affects a huge number of recordings in MB, including just about anything released before 1982 and quite a bit that was released after.

If the music is essentially the same other than what we might call "noise", then they should be considered the same, in my opinion. I know some believe that any audible difference, no matter how slight, makes recordings different, but I don't think we should be that strict. Any audiophile understands that there are differences between vinyl, CD, cassette etc. and will make allowances. It will just make more work, and so reduce the depth of the database, if we also start breaking up recordings along those same distinctions.

If we do want to get down to this level of detail, then we should do some work first in the schema and web site to avoid redundancy and "make work" activities where possible. For example:

- On the web site, and in the API, link a remastered recording back to the relationships of the earlier version, so that we can avoid re-entering them all. We might want to skip specific relationships e.g. the 'mastered by' if they don't make sense, but things like performance relationships that apply equally to both should be inherited.

- Make it easy to create remastered recordings at the release level, since we'll be doing a lot of that to convert over all of the releases that have both Vinyl / Cassette and CD / Digital / DAT / ... formats. At present the tracklist editor just lets you 'add a new recording' which leaves a lot of work to do in linking those recordings to the earlier analogue (or whatever) versions. We could get a bot to do this, but it will be an ongoing activity as studios release digitized versions of old albums.

- Add a 'Digitized' relationship like the 'remastering' relationship, which would allow digital recordings to be linked to a corresponding analogue. This could be generalized to some kind of 'format change' relationship. In practice, though, I suspect most digital releases were taken from the original masters and not simple digitizations, so would count as 'remasters' (though getting this info from a studio is likely problematic).

This is my first Style Proposal, so if I've done part of it incorrectly or I've missed some part of the procedure, please let me know! However, I think everything is as it should be.

Kind Regards,
Ben (LordSputnik)


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
Splitting Vinyl and CD recordings is a lot of work, but in most cases I doubt the difference is just noise, at least for remasters made in the last 20 years. Perhaps some early CDs from the 80s were just direct transfers, such as the early 80s ABBA CD releases, for example.

However, these cases are probably few. It seems fundamentally wrong to me for vinyls to share recordings with CDs, since they aren't in any way related, and the audio will definitely change in a remaster. And if you look at the IRC log from last week, that's the very topic that started off this proposal!

I wouldn't go so far as to split digitization and remastering - I think digitizing should count as remastering, since the master is being transferred from one type to another.

A lot of your ideas about making it easier to designate remastered recordings are good ones. If this style proposal succeeds, you should definitely make tickets for them.

On 31 October 2012 19:23, Sheamus Patt <[hidden email]> wrote:
On 10/31/2012 01:29 PM, Ben Ockmore wrote:
Expected Expiration Date for RFC: Wednesday, 7th November 2012


This proposal is designed to improve the guidelines on merging recordings, at https://wiki.musicbrainz.org/Style/Recording.

It is felt that the current guidelines are too vague, and that they don't cover all cases where recordings might be merged. This proposal seeks to expand these guidelines and make them clearer.

JIRA Ticket (includes previous IRC discussion): http://tickets.musicbrainz.org/browse/STYLE-159
Wiki Page: https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines


I think it's going too far to consider recordings different simply because one is analogue (e.g. Vinyl) and the other digital (e.g. CD). This affects a huge number of recordings in MB, including just about anything released before 1982 and quite a bit that was released after.

If the music is essentially the same other than what we might call "noise", then they should be considered the same, in my opinion. I know some believe that any audible difference, no matter how slight, makes recordings different, but I don't think we should be that strict. Any audiophile understands that there are differences between vinyl, CD, cassette etc. and will make allowances. It will just make more work, and so reduce the depth of the database, if we also start breaking up recordings along those same distinctions.

If we do want to get down to this level of detail, then we should do some work first in the schema and web site to avoid redundancy and "make work" activities where possible. For example:

- On the web site, and in the API, link a remastered recording back to the relationships of the earlier version, so that we can avoid re-entering them all. We might want to skip specific relationships e.g. the 'mastered by' if they don't make sense, but things like performance relationships that apply equally to both should be inherited.

- Make it easy to create remastered recordings at the release level, since we'll be doing a lot of that to convert over all of the releases that have both Vinyl / Cassette and CD / Digital / DAT / ... formats. At present the tracklist editor just lets you 'add a new recording' which leaves a lot of work to do in linking those recordings to the earlier analogue (or whatever) versions. We could get a bot to do this, but it will be an ongoing activity as studios release digitized versions of old albums.

- Add a 'Digitized' relationship like the 'remastering' relationship, which would allow digital recordings to be linked to a corresponding analogue. This could be generalized to some kind of 'format change' relationship. In practice, though, I suspect most digital releases were taken from the original masters and not simple digitizations, so would count as 'remasters' (though getting this info from a studio is likely problematic).

This is my first Style Proposal, so if I've done part of it incorrectly or I've missed some part of the procedure, please let me know! However, I think everything is as it should be.

Kind Regards,
Ben (LordSputnik)


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Alex Mauer
In reply to this post by Sheamus Patt
On 10/31/2012 02:23 PM, Sheamus Patt wrote:

> On 10/31/2012 01:29 PM, Ben Ockmore wrote:
>> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>>
>>
>> This proposal is designed to improve the guidelines on merging
>> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>>
>> It is felt that the current guidelines are too vague, and that they
>> don't cover all cases where recordings might be merged. This proposal
>> seeks to expand these guidelines and make them clearer.
>>
>> JIRA Ticket (includes previous IRC discussion):
>> http://tickets.musicbrainz.org/browse/STYLE-159
>> Wiki Page:
>> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>>
>>
> I think it's going too far to consider recordings different simply
> because one is analogue (e.g. Vinyl) and the other digital (e.g. CD).
> This affects a huge number of recordings in MB, including just about
> anything released before 1982 and quite a bit that was released after.
The problem with considering them the same, is when multiple
analog→digital transfers happen at different times using different
technologies, possibly from different sources (common with old 78RPM
records: the original master is often lost/destroyed and different
physical copies are used for the remaster source.) Often these will be
audibly different from each other, despite all being taken from what is
ostensibly “the same” master. And if we decide that one of them is
allowed to be the same, how do we decide which one gets to be that
“special” one? Easier just to say “digital cannot be the same as analog”
and leave it at that.

> - On the web site, and in the API, link a remastered recording back to
> the relationships of the earlier version, so that we can avoid
> re-entering them all. We might want to skip specific relationships e.g.
> the 'mastered by' if they don't make sense, but things like performance
> relationships that apply equally to both should be inherited.

I know I have kicked around the idea of “recording groups” (analogous to
release groups) which would allow for this. Regardless of what they
might be called, I think this would be a very good idea.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style

signature.asc (563 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

swisschris
Having three different "MB-Recordings" for a track issued simultaneously on vinyl, CD and cassette (quite common throughout the eighties) seems absurd – or at least counter-intuitive. And what about digital download vs physical (CD) Media? Two different recordings as well? And different whether downloaded from amazon or iTunes? Different whether downloaded as MP3 or WAV or AIFF?

If guidelines define 19 out of 20 tracks as unique and thus un-mergeable, what's the use of "MB-Recordings"?

On Wed, Oct 31, 2012 at 9:01 PM, Alex Mauer <[hidden email]> wrote:
On 10/31/2012 02:23 PM, Sheamus Patt wrote:
> On 10/31/2012 01:29 PM, Ben Ockmore wrote:
>> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>>
>>
>> This proposal is designed to improve the guidelines on merging
>> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>>
>> It is felt that the current guidelines are too vague, and that they
>> don't cover all cases where recordings might be merged. This proposal
>> seeks to expand these guidelines and make them clearer.
>>
>> JIRA Ticket (includes previous IRC discussion):
>> http://tickets.musicbrainz.org/browse/STYLE-159
>> Wiki Page:
>> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>>
>>
> I think it's going too far to consider recordings different simply
> because one is analogue (e.g. Vinyl) and the other digital (e.g. CD).
> This affects a huge number of recordings in MB, including just about
> anything released before 1982 and quite a bit that was released after.

The problem with considering them the same, is when multiple
analog→digital transfers happen at different times using different
technologies, possibly from different sources (common with old 78RPM
records: the original master is often lost/destroyed and different
physical copies are used for the remaster source.) Often these will be
audibly different from each other, despite all being taken from what is
ostensibly “the same” master. And if we decide that one of them is
allowed to be the same, how do we decide which one gets to be that
“special” one? Easier just to say “digital cannot be the same as analog”
and leave it at that.

> - On the web site, and in the API, link a remastered recording back to
> the relationships of the earlier version, so that we can avoid
> re-entering them all. We might want to skip specific relationships e.g.
> the 'mastered by' if they don't make sense, but things like performance
> relationships that apply equally to both should be inherited.

I know I have kicked around the idea of “recording groups” (analogous to
release groups) which would allow for this. Regardless of what they
might be called, I think this would be a very good idea.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Sheamus Patt
In reply to this post by Alex Mauer
On 10/31/2012 04:01 PM, Alex Mauer wrote:
On 10/31/2012 02:23 PM, Sheamus Patt wrote:
On 10/31/2012 01:29 PM, Ben Ockmore wrote:
Expected Expiration Date for RFC: Wednesday, 7th November 2012


This proposal is designed to improve the guidelines on merging
recordings, at https://wiki.musicbrainz.org/Style/Recording.

...
I think it's going too far to consider recordings different simply
because one is analogue (e.g. Vinyl) and the other digital (e.g. CD).
This affects a huge number of recordings in MB, including just about
anything released before 1982 and quite a bit that was released after.
The problem with considering them the same, is when multiple
analog→digital transfers happen at different times using different
technologies, possibly from different sources (common with old 78RPM
records: the original master is often lost/destroyed and different
physical copies are used for the remaster source.) Often these will be
audibly different from each other, despite all being taken from what is
ostensibly “the same” master. And if we decide that one of them is
allowed to be the same, how do we decide which one gets to be that
“special” one? Easier just to say “digital cannot be the same as analog”
and leave it at that.
I don't dispute that there are audible differences, but to me, they are just academic; for the purposes of documenting the content and heritage of recorded music, I just don't care. I don't really see just how this adds value to most MusicBrainz users either. Can you explain how you would use this information, beyond what's already available from the Release Format (e.g. CD or 12" Vinyl)? After all, you can't actually listen to the music on MusicBrainz.

I could go along with splitting up recordings based on format, provided that it's easy to use and doesn't weigh down editors with the effort of duplicating release-artist relationships on each copy. I find those relationships very informative - I want to know who's playing and singing background on tracks I'm listening to. Whether a chorus got repeated on the recording on a particular release, but not on the "same" recording on some other Various Artists use of that track - not so much! We already see this same sort of issue with releases and release-groups. If you need to add relationships to releases, they you likely need to duplicate them on all of the releases in the same release group, and this usually doesn't get done.

Incidentally, the current release editor / server has bugs e.g. http://tickets.musicbrainz.org/browse/MBS-3307 that make this kind of splitting even more difficult. I recently needed to split off a recording, in this case because it never should have been merged http://musicbrainz.org/recording/ee7f4e9c-353f-445c-9001-37bbe3c4bd57 . The problem is that when you use the Recordings editor to "Add a New Recording", it seems that the new recording isn't accessible until the edit is vetted, usually in 2 weeks (and I'm just assuming). See, for example, this edit on the test server http://test.musicbrainz.org/release/ec6f2c21-0927-44c3-85a8-6949fa8cc65a/open_edits which lets you access the old recording, but not the new. My workaround is not likely one we want to use a lot; I created a Standalone Recording first, and substituted that one. If we're going to be splitting off recordings from entire releases as remasters, then we need a better way to do this.



- On the web site, and in the API, link a remastered recording back to
the relationships of the earlier version, so that we can avoid
re-entering them all. We might want to skip specific relationships e.g.
the 'mastered by' if they don't make sense, but things like performance
relationships that apply equally to both should be inherited.
I know I have kicked around the idea of “recording groups” (analogous to
release groups) which would allow for this. Regardless of what they
might be called, I think this would be a very good idea.

That could work. I don't think we'd want a group for each recording, so it should only be created when you need to associate several. For that reason I think the idea of "inheriting" relationships would be a bit cleaner, but the end result wouldn't be much different. A lot depends on how usable the end result would be.




_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
In reply to this post by swisschris
You wouldn't want three separate recordings. That would only be two. One for each audio source used.

The cassette and vinyl would both probably use the original analogue master as their audio source, so they would share recordings. The CD would use a digitzed audio source, so it would need another recording.

Downloads released at the same time as a CD, or ripped from a CD are all still based on the digital audio source, so they all share recordings. The only differences in the audio there are due to encoding issues, which aren't going to result in significant differences in the audio and don't affect the original data of the audio source.

On 31 October 2012 21:21, SwissChris <[hidden email]> wrote:
Having three different "MB-Recordings" for a track issued simultaneously on vinyl, CD and cassette (quite common throughout the eighties) seems absurd – or at least counter-intuitive. And what about digital download vs physical (CD) Media? Two different recordings as well? And different whether downloaded from amazon or iTunes? Different whether downloaded as MP3 or WAV or AIFF?

If guidelines define 19 out of 20 tracks as unique and thus un-mergeable, what's the use of "MB-Recordings"?

On Wed, Oct 31, 2012 at 9:01 PM, Alex Mauer <[hidden email]> wrote:
On 10/31/2012 02:23 PM, Sheamus Patt wrote:
> On 10/31/2012 01:29 PM, Ben Ockmore wrote:
>> Expected Expiration Date for RFC: Wednesday, 7th November 2012
>>
>>
>> This proposal is designed to improve the guidelines on merging
>> recordings, at https://wiki.musicbrainz.org/Style/Recording.
>>
>> It is felt that the current guidelines are too vague, and that they
>> don't cover all cases where recordings might be merged. This proposal
>> seeks to expand these guidelines and make them clearer.
>>
>> JIRA Ticket (includes previous IRC discussion):
>> http://tickets.musicbrainz.org/browse/STYLE-159
>> Wiki Page:
>> https://wiki.musicbrainz.org/User:LordSputnik/Recording_Style_Guidelines
>>
>>
> I think it's going too far to consider recordings different simply
> because one is analogue (e.g. Vinyl) and the other digital (e.g. CD).
> This affects a huge number of recordings in MB, including just about
> anything released before 1982 and quite a bit that was released after.

The problem with considering them the same, is when multiple
analog→digital transfers happen at different times using different
technologies, possibly from different sources (common with old 78RPM
records: the original master is often lost/destroyed and different
physical copies are used for the remaster source.) Often these will be
audibly different from each other, despite all being taken from what is
ostensibly “the same” master. And if we decide that one of them is
allowed to be the same, how do we decide which one gets to be that
“special” one? Easier just to say “digital cannot be the same as analog”
and leave it at that.

> - On the web site, and in the API, link a remastered recording back to
> the relationships of the earlier version, so that we can avoid
> re-entering them all. We might want to skip specific relationships e.g.
> the 'mastered by' if they don't make sense, but things like performance
> relationships that apply equally to both should be inherited.

I know I have kicked around the idea of “recording groups” (analogous to
release groups) which would allow for this. Regardless of what they
might be called, I think this would be a very good idea.


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

LordSputnik
In reply to this post by Sheamus Patt
Not having separate recordings for analogue and digital masters is misleading. They aren't the same recording - they have different audio, and often that difference is noticeable. If we have different recordings for remasters, we should have different recordings for analogue and digital masters.

I agree that a better way of fixing remasters is needed. I've been splitting the Beatles mono and stereo remasters recently, and it's taken me about a month so far to do 13 mediums...

On 31 October 2012 21:36, Sheamus Patt <[hidden email]> wrote:
On 10/31/2012 04:01 PM, Alex Mauer wrote:
On 10/31/2012 02:23 PM, Sheamus Patt wrote:
On 10/31/2012 01:29 PM, Ben Ockmore wrote:
Expected Expiration Date for RFC: Wednesday, 7th November 2012 This proposal is designed to improve the guidelines on merging recordings, at https://wiki.musicbrainz.org/Style/Recording
. ...
I think it's going too far to consider recordings different simply
because one is analogue (e.g. Vinyl) and the other digital (e.g. CD).
This affects a huge number of recordings in MB, including just about
anything released before 1982 and quite a bit that was released after.
The problem with considering them the same, is when multiple
analog→digital transfers happen at different times using different
technologies, possibly from different sources (common with old 78RPM
records: the original master is often lost/destroyed and different
physical copies are used for the remaster source.) Often these will be
audibly different from each other, despite all being taken from what is
ostensibly “the same” master. And if we decide that one of them is
allowed to be the same, how do we decide which one gets to be that
“special” one? Easier just to say “digital cannot be the same as analog”
and leave it at that.
I don't dispute that there are audible differences, but to me, they are just academic; for the purposes of documenting the content and heritage of recorded music, I just don't care. I don't really see just how this adds value to most MusicBrainz users either. Can you explain how you would use this information, beyond what's already available from the Release Format (e.g. CD or 12" Vinyl)? After all, you can't actually listen to the music on MusicBrainz.

I could go along with splitting up recordings based on format, provided that it's easy to use and doesn't weigh down editors with the effort of duplicating release-artist relationships on each copy. I find those relationships very informative - I want to know who's playing and singing background on tracks I'm listening to. Whether a chorus got repeated on the recording on a particular release, but not on the "same" recording on some other Various Artists use of that track - not so much! We already see this same sort of issue with releases and release-groups. If you need to add relationships to releases, they you likely need to duplicate them on all of the releases in the same release group, and this usually doesn't get done.

Incidentally, the current release editor / server has bugs e.g. http://tickets.musicbrainz.org/browse/MBS-3307 that make this kind of splitting even more difficult. I recently needed to split off a recording, in this case because it never should have been merged http://musicbrainz.org/recording/ee7f4e9c-353f-445c-9001-37bbe3c4bd57 . The problem is that when you use the Recordings editor to "Add a New Recording", it seems that the new recording isn't accessible until the edit is vetted, usually in 2 weeks (and I'm just assuming). See, for example, this edit on the test server http://test.musicbrainz.org/release/ec6f2c21-0927-44c3-85a8-6949fa8cc65a/open_edits which lets you access the old recording, but not the new. My workaround is not likely one we want to use a lot; I created a Standalone Recording first, and substituted that one. If we're going to be splitting off recordings from entire releases as remasters, then we need a better way to do this.




      
- On the web site, and in the API, link a remastered recording back to
the relationships of the earlier version, so that we can avoid
re-entering them all. We might want to skip specific relationships e.g.
the 'mastered by' if they don't make sense, but things like performance
relationships that apply equally to both should be inherited.
I know I have kicked around the idea of “recording groups” (analogous to
release groups) which would allow for this. Regardless of what they
might be called, I think this would be a very good idea.

That could work. I don't think we'd want a group for each recording, so it should only be created when you need to associate several. For that reason I think the idea of "inheriting" relationships would be a bit cleaner, but the end result wouldn't be much different. A lot depends on how usable the end result would be.


      

_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Alex Mauer
In reply to this post by Sheamus Patt
On 10/31/2012 04:36 PM, Sheamus Patt wrote:
> I don't dispute that there are audible differences, but to me, they are
> just academic; for the purposes of documenting the content and heritage
> of recorded music, I just don't care. I don't really see just how this
> adds value to most MusicBrainz users either. Can you explain how you
> would use this information, beyond what's already available from the
> Release Format (e.g. CD or 12" Vinyl)? After all, you can't actually
> listen to the music on MusicBrainz.

Academic to you, but not to everyone. For example, someone studying the
Loudness War would have interest in tracking the changes of different
remasters over time.

> If you need to add relationships to releases, they you likely need to
> duplicate them on all of the releases in the same release group, and
> this usually doesn't get done.

Yes, this can be a problem. The solution is to create better tools; the
new relationship editor is a good start in this direction.
>> I know I have kicked around the idea of "recording groups" (analogous to
>> release groups) which would allow for this. Regardless of what they
>> might be called, I think this would be a very good idea.
>
> That could work. I don't think we'd want a group for each recording, so
> it should only be created when you need to associate several. For that
> reason I think the idea of "inheriting" relationships would be a bit
> cleaner, but the end result wouldn't be much different. A lot depends on
> how usable the end result would be.

I agree. As I said, it’s just an idea at the moment. Implementation
would nail down the details of how it might work in practice. (and even
once it’s implemented, the tools can always be improved)



_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style

signature.asc (563 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Frederic Da Vitoria
2012/10/31 Alex Mauer <[hidden email]>
On 10/31/2012 04:36 PM, Sheamus Patt wrote:
> I don't dispute that there are audible differences, but to me, they are
> just academic; for the purposes of documenting the content and heritage
> of recorded music, I just don't care. I don't really see just how this
> adds value to most MusicBrainz users either. Can you explain how you
> would use this information, beyond what's already available from the
> Release Format (e.g. CD or 12" Vinyl)? After all, you can't actually
> listen to the music on MusicBrainz.

Academic to you, but not to everyone. For example, someone studying the
Loudness War would have interest in tracking the changes of different
remasters over time.

Not only someone studying the Loudness Wars, someone who simply would want to recover the sound he remembers from before the recording was "digitally enhanced" :-(

There are tracks for which I would definitely like to know if their dynamics were "compressed" or not, and I hope someday MB will help me find them.

--
Frederic Da Vitoria
(davitof)

Membre de l'April - « promouvoir et défendre le logiciel libre » - http://www.april.org


_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

lorenz pressler

2012/10/31 Alex Mauer <[hidden email]>

> Academic to you, but not to everyone. For example, someone studying the
> Loudness War would have interest in tracking the changes of different
> remasters over time.

isn't that the meaning of academic?



Frederic Da Vitoria <[hidden email]>:

> [...] someone who simply would want
> to recover the sound he remembers from before the recording was  
> "digitally
> enhanced" :-(
> There are tracks for which I would definitely like to know if their
> dynamics were "compressed" or not, and I hope someday MB will help me  
> find them.

will it be easier to identify a specific quality of a recording if there  
are a lot more MB-recordings around? would we want to store the loudness  
factor for recordings? we would still need some (subjective!) quality  
rating for different remasters to actually gain some info that has some  
real life use.

personally i'd love a database that has complete (re)master/transfer  
information about every release, however i have not the slightest idea  
where this information should come from. of all the releases i actually  
own only a handful provide information about what sources were used. of  
course, if you edit some beatles recordings you will find plenty of stuff,  
lots remasters and lots of information about these remasters. but just  
because there are some small fields in the database that work out well  
with the proposal this might not be true for the majority of the data.  
more helpful would be linking reviews of a specific release that covers  
some discussion about the quality of the remastering. from my personal  
experience the evaluation of the quality of mastering job is highly  
subjective and i've seen fierce flamewars about what remaster is 'better'.


ad 3:
> If a CD or other digital release contains audio that was originally  
> recordedand released in an analogue format, the digital release should  
> be given anew set of recordings. The original analogue source was  
> changed when it wasconverted from analogue to digital.

afaik pre-NGS releases now share recordings between CD and vinyl versions.  
theses would have to be split and all the metadata recreated. i just see a  
lot of work, and no real gain. if i look at any release pre-80s it is  
obvious that the later released CD version will be a transfer.


ad 6:
> If a release is remastered from an earlier release, all the tracks on  
> the newrelease should be new recordings. The source was changed when the  
> recordingwas remastered. If another remaster is made, then the source  
> has changedagain, and another set of recordings should be made.

i've not seen many remasters that were worth mentioning. imho only  
recordings that have some information about the remastering process  
published or when there are sourced discussions about differences then it  
should be allowed to have separate recordings. having different recordings  
just for the sake of it is a bad idea. we would need catch-all recording  
for compilations and releases that don't give information about (re)master  
information and loose the possibility to share metadata.

for me this proposal is a regression. as nice it may sound i don't think  
this is manage- and editable.


best, lorenz.

_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Reply | Threaded
Open this post in threaded view
|

Re: RFC: Improve Guidelines on Merging Recordings

Alex Mauer
On 10/31/2012 9:58 PM, lorenz pressler wrote:
>
> 2012/10/31 Alex Mauer <[hidden email]>
>
>> Academic to you, but not to everyone. For example, someone studying the
>> Loudness War would have interest in tracking the changes of different
>> remasters over time.
>
> isn't that the meaning of academic?

It's *a* meaning of academic, yes. I took it as meaning "Not of
practical relevance; of only theoretical interest" since Sheamus was
using the term dismissively.

> Frederic Da Vitoria <[hidden email]>:
>
>> [...] someone who simply would want
>> to recover the sound he remembers from before the recording was
>> "digitally
>> enhanced" :-(
>> There are tracks for which I would definitely like to know if their
>> dynamics were "compressed" or not, and I hope someday MB will help me
>> find them.
>
> will it be easier to identify a specific quality of a recording if there
> are a lot more MB-recordings around? would we want to store the loudness
> factor for recordings? we would still need some (subjective!) quality
> rating for different remasters to actually gain some info that has some
> real life use.

In order to even think about deciding which masters/remasters might be
better or best...we need to have them separated. If we can't even
identify the *different* recordings, we certainly can't think about
comparing them accurately, or enabling people to find the ones that are
"better" if we do.

> ad 3:
>> If a CD or other digital release contains audio that was originally
>> recordedand released in an analogue format, the digital release should
>> be given anew set of recordings. The original analogue source was
>> changed when it wasconverted from analogue to digital.
>
> afaik pre-NGS releases now share recordings between CD and vinyl versions.
> theses would have to be split and all the metadata recreated. i just see a
> lot of work, and no real gain. if i look at any release pre-80s it is
> obvious that the later released CD version will be a transfer.

And what about a different later released CD version? Is that a new
transfer, or a copy of the original transfer? Or an edit of the original
transfer?

> ad 6:
>> If a release is remastered from an earlier release, all the tracks on
>> the newrelease should be new recordings. The source was changed when the
>> recordingwas remastered. If another remaster is made, then the source
>> has changedagain, and another set of recordings should be made.
>
> i've not seen many remasters that were worth mentioning. imho only
> recordings that have some information about the remastering process
> published or when there are sourced discussions about differences then it
> should be allowed to have separate recordings. having different recordings
> just for the sake of it is a bad idea. we would need catch-all recording
> for compilations and releases that don't give information about (re)master
> information and loose the possibility to share metadata.

You might want to re-read the current Recording style guidelines -- they
basically already say all this stuff:
http://musicbrainz.org/doc/Style/Recording ; the new proposed guidelines
just make it clearer and read better.



_______________________________________________
MusicBrainz-style mailing list
[hidden email]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
1234 ... 6