No kerning for MuseJazzText font due to lost SFD file
Reported version
3.0
Priority
P1 - High
Type
Development
Frequency
Once
Severity
S3 - Major
Reproducibility
Always
Status
closed
Regression
Yes
Workaround
No
Project
In 2.x we had a MuseJazz font (in normal and Macintosh flavours) and its .sfd source code.
In 3.x/master we have MuseJazz and MuseJazzText, both of which lack the corresponding source code.
Unless there is no FontForge source file (.sfd) for these any more and the .otf files are now used as “source” when loading the font into FontForge for editing (which I don’t believe), the .sfd files MUST be added back to the repository, or we cannot ship either font in Debian.
Cross-reference: https://github.com/musescore/MuseScore/commit/86b27e570ddc6827798918e59…
Fix version
3.6.0
Comments
I'm afraid we don't have sfd anymore for these fonts. I don't even know if the editing was done in FontForge. Also, note that MuseJazz in MuseScore 2 was a text font and so now is closer from MuseJazzText while MuseJazz is a score font now on MuseScore 3.
anyone remembers who created this new MuseJazz or where it came from?
Edit: 86b27e5 and 83e1d3c
In reply to anyone remembers who created… by Jojo-Schmitz
@lasconic: so you now all open the
.otffiles in order to edit the font, including the designer?@Jojo: according to https://musescore.org/en/node/137146 at least, Abraham Lee from Music Type Foundry has created the MuseJazz musical font (although nowhere is declared whether it’s a derived work of something).
The MuseJazzText font is apparently a derivative of the old (wschweer’s) MuseJazz font, also by Abraham Lee (?), but its licencing information lacks this:
MuseJazz (old, text font) self-identifies as “MuseJazz-Book” (where “-Book” is a style, not part of the font name) and says: “Created by Werner Schweer” and “Copyright (c) 2009, Werner Schweer,,, (), with Reserved Font Name MuseJazz.”
MuseJazzText (new, text font) self-identifies as “MuseJazzText” and says: “Copyright (c) 2014-2016 by Abraham Lee. Licensed under SIL OFL.” and “Copyright © 2016, Abraham Lee (https://www.musictypefoundry.com/), with Reserved Font Name "MuseJazz".” (The “TTF Names” part of the font metadata shows more inconsistencies looking to come from an incompatible rename.)
MuseJazz (new, musical font) self-identifies as “MuseJazz” and says: “Copyright (c) 2014-2016 by Abraham Lee. Licensed under SIL OFL.” and “Copyright © 2016, Abraham Lee (https://www.musictypefoundry.com/), with Reserved Font Name "MuseJazz".”
So, at the very least, we have three distinct fonts that use the same SIL OFL “Reserved Font Name”, one of them even using a different font name. Reuse of the Reserved Font Name is only acceptable with permission from the one originally reserving it (wschweer), but I assume this either exists or is trivial to get or is part of “schlüssiges Verhalten” (by accepting the commit into the repo).
Furthermore, MuseJazzText is apparently a derivative of MuseJazz-Book (the old one) but does not retain its copyright line. This is not good and should be raised with Abraham Lee.
MuseJazz (the new musical font) contains symbols that also occur in MuseJazzText. Iff these are a derivative work of the old MuseJazz-book font (which also did contain some musical symbols), the same applies here. (Ask Abraham Lee whether he copied anything from old MuseJazz-Book to new MuseJazz.)
As for the sources, this is also a question I’d pose to Abraham Lee: whether, when he were to edit either of the two new fonts right now, he’d open the
.otffiles in his editor, or has any other “source” files.Just to clarify: if everyone were to treat the
.otffiles as source and there existed no other files that could be considered source, that’s acceptable to Debian. It’s just that we used to have accompanying.sfdfiles and now don’t, which raises red flags.As discussed at FOSDEM:
MuseScore upstream does not have anything other than the OTF files; no SFD source exists any more, and the external foundry probably does not use FontForge. If MuseScore upstream were to publish a changed version, they’d start from the OTF files, which are lossless enough for this. This is sufficient for my definition of corresponding source code as required by Debian, and therefore I’ll remove the exclusion of MuseJazz{, Text} and allow it into Debian with just the OTF files present.
In reply to As discussed at FOSDEM: … by mirabilos
One question remains: is the new MuseJazz Text a derivative work of the old MuseJazz? (And possibly the new MuseJazz, as it also ships text glyphs.)
If so, the
© 2009 Werner Schweeris missing from it.Somehow I missed this discussion, sorry.
We certainly still have the 2.3.2 SFD files for MuseJazz, which is now called MuseJazzText. I can't say if they are still current - when the change from MuseJazz to MuseJazzText happened, I suspect some glyphs were rearranged, renamed, etc. But what I can say is that along the way, we lost the kerning info from MuseJazzText, and that's a significant regression. So if nothing else, we should try to copy that info over from the old MuseJazz.sfd and apply it to the current MuseJazzText. The thing is, I don't really know how to do that.
In reply to Somehow I missed this… by Marc Sabatella
As far as I could figure out from the
.otf, MuseJazzText is not just 2.x MuseJazz either, but a new font created from parts of the 2.x MuseJazz and parts of the 3.x MuseJazz, or something. It certainly bears Abraham Lee’s copyright… but is lacking Werner’s, which is wrong if it’s really a derivative of the 2.x MuseJazz. (If it’s just the 2.x MuseJazz, then Abraham Lee’s copyright is actually copyfraud…)In the Debian licencing metadata, I’ve “solved” this by stating the directory’s contents are under OFL, with Abraham Lee’s copyright and Werner’s added as “maybe”. Not very satisfying, but apparently nobody knows what really happened with these fonts. They certainly look close enough to the 2.x MuseJazz to be a derivative, but I’m sure nobody wants to sue Abraham Lee either and thus disgruntle him…
OK, with the information received today, I was able to update our metadata in Debian:
https://evolvis.org/plugins/scmgit/cgi-bin/gitweb.cgi?p=alioth/musescor…
With this, and the info I got at FOSDEM, I consider this closed from the Debian side, although it would be safer to have the identical information in the
.otffiles to appreciate all authors (after all, a CLA is not a copyright assignment (which is not possible in most of Europe anyway) but a very broad licence, so the authors have a right to be named), I will not insist on that.So, feel free to either close this issue directly or take it as TODO to update the
.otfsome day (perhaps next time they’ll be updated anyway, to save a roundtrip through fontforge).Hm, one more thing… Marc, if any of the glyphs you created are in there, you also have a right to be named ☺ please update the information accordingly. (Perhaps I should add you to the 2.3.2 copy of the debian/copyright file… with which year(s)?)
The lack of kerning info is still a real issue directly attributed to the loss of the SFD file, so I'd prefer to leave this open. Or we could open a new issue specifically about kerning, but since it really is about the SFD, I'd just leave this one open.
As for my glyphs, I forget how many I did, but frankly I'd be happier to see mine removed and replaced by better ones. Like, my segno and coda are still there, but so are ones that I assume Abraham Lee designed. I did the kerning, but it's gone... Anyhow, I don't see any reason to associate my name with this file any more than the many other source files I worked on, so I think I'll pass :-)
Now that I'm dealing FontForge a bit more because of my work with Campania, I have ideas on how to repair this. I should be able to started from the 2.3.2 version, have FontForge generate feature files (which should encapsulate the kerning), edit those as needed, and merge those in the current OTF. I could also save the resulting SFD and commit it - I see no reason we shouldn't have a source for this. Unless there is some legal reason not to. Whatever, I just want to fix the kerning.
The comment about the "Reserved name" conflict is possibly relevant to #279563: [MacOS X] Export Pdf with Muse Jazz Text all garbled and unreadable if no printers are setup on a machine, though.
No, the “Reserved name” is just a thing specific to the SIL OFL licence. It basically says “if you change the font and are not the original author, you must rename it and cannot keep the reserved name”.
In reply to No, the “Reserved name” is… by mirabilos
I finally got around to transferring the kern info from the 2.3.2 version. Unfortunately, it produces terrible results - I think somewhere between then and now, we also narrowed a bunch of the bearings. However, the autokern feature of FontForge does a passable job, so I've made a pass through the font using that. It's not completely automatic, you still have to decide which classes of characters to combine etc.
It seems going forward we may not be using FontForge anymore anyhow, so I believe I will simply update the OTF file.
see https://github.com/musescore/MuseScore/pull/7091
https://github.com/musescore/MuseScore/pull/7091
Fixed in branch 3.x, commit 9014df4874
_fix #277756: no kerning for MuseJazzText
Resolves: https://musescore.org/en/node/277756
When MuseJazzText was first introduced (as MuseJazz),
there was quite a bit of good kerning info.
This was lost when MuseJazzText was split of from MuseJazz -
the kern tables never made the transition.
Although the old kern information is still available in GitHub,
it turns out to no longer work well - too much has changed in the font
(in particu,ar, it seems glyph bearings have been narrowed).
So this commit creates a brand new kern table to MuseJazzText.otf,
using FontForge's autokern feature to generate some reasonable values
for a hand-selected set of character classes._
Automatically closed -- issue fixed for 2 weeks with no activity.