Dave Crossland wrote: > Hi, > > You can download a chat between Ed Trager, Ben Weiner and I about the > development of the site we had tonight - its 31Mb at > http://www.openfontlibrary.org/2009-05-31_OFLB_devtalk.mp3
Great to know some more about the ongoing work and the components coming
together!
Here's some feedback after listening to the discussion:
I think the consensus on the upload requirements are great and provide a
reasonable way of sending the "we care about quality" message, not
requiring too much of designers to get their buy-in but still making
long-term maintainership of the library possible.
A few comments though:
- for the FONTLOG: the recommended form is FONTLOG.txt (also used as
such by the majority of open fonts I've packaged or seen released so
far), we can take advantage of your regexps to rename to that at the
upload stage.
- how do we handle families/weight variants: foo-1.3-regular.ttf
foo-1.3-italic.ttf etc? We say at least one .ttf? we create a record
for each font and a meta-record for the family?
- the LICENSE file: I agree that we don't want files without a license!
But for many open fonts licenses the LICENSE file is actually OFL.txt,
MIT.txt, or something like GPL-3-font-exception.txt. We should take
these into account. If the metadata in the font itself has empty
copyright/license/licenseURL fields we should reject it too. The
external license files should properly mirror what is in the ttf
metadata fields. One without the other is not good.
So we'd accept tarballs and zips structured like
foo-1.3/FONTLOG.txt
/foo-1.3.ttf (or foo-1.3-regular.ttf / foo-1.3-bold.ttf)
/OFL.txt | MIT.txt | GPL-3-font-exception.txt
Also the uploading mechanism should strip away any hidden files like
.DS_store or similar (.svn/ .hg / .bzr |.git / whatever) which may find
themselves left behind in the release tarball.
About your example of Vietnamese LICENSE file: we must be aware that
translated licenses are only unofficial, we need to require the original
license file written in eng. In our accepted spectrum of licenses it's
always the English version which is the original. (for example we
currently have a unofficial vie version of the OFL, dunno about the others).
I very much agree with a page in the site for people to test out their
metadata/coverage/etc (well OFLB policy compliance actually) separated
from the full upload process. An author would use that page to test, get
reports, do the needed mods, re-test, realize it's good then go to the
upload page.
I also agree with making publication of extended font sources optional
(like .sfd, .fea, .gdl, .ttx, .xgf, .vtp, .aat, .xml, .xsl, .ufo/,
.sfdir/, etc. and build-related items) the FONTLOG can always refer to
them and we can add a mention to that effect in our upload
documentation: the more extended font sources published the better but
it's really the author's choice.
I fully agree that it makes good sense to be inclusive and not strictly
require usage of the open font design toolkit.
The uploaded fonts should not be fully published before going through a
quick manual review process from one of the library admins. A light
moderation pipeline is needed: we haven't been capable of doing
post-moderation so far.
Btw, the SIL fonts are also developed, reviewed and post-processed with
perl/python scripts / libraries published for others to use:
http://scripts.sil.org/FontUtils, and there is also the Graphite
compiler, Reprise, TypeTuner, etc. Let me say it again: more of these
libraries, tools and corresponding documentation are being prepared for
release to the community as well: it takes time and effort. And many
designers and script engineers are watching the open font toolkit as it
nicely matures :-)
The pronunciation of Gentium is indicated in the FAQ:
The preferred pronunciation is with a soft G as in 'general', not a
hard one as in 'gold': JEN-tee-oom.
I agree that the fontlinking URI should be unique to a "branch", when a
new modification is done it's a new font, a new branch, a new address.
It should include the name of the family and the version. Users will cut
and paste such addresses to their CSS anyway, it's up to us not to mess
up the namespaces of the various branches.
We really need a clear version numbering system like:
$fontfamilyname-$major.$minor-$OFLB-release-id
$foo-1.3-1
Once a font is uploaded, an OFLB-specific release-id (typeface record)
is added to the upstream major.minor.
The fontlinking URI should then be
http://openfontlibrary.org/people/authorname/foo-1.3-1.ttf
An apache mod_rewrite rule can handle that if it's tricky with cchost.
When a version is published: the corresponding URI stays, cool URIs
don't change. And we put description items indicating the corresponding
branches in the pages presenting each font.
This takes into account that the OFLB is a library not an upstream
hosting site: OFLB publication is an end-stage of upstreams, upstream
will release new versions which will be packaged on the desktops as well
as released on the library for web use: the number versionning should
support that and not create conflicts. Branches should be
distinguishable among themselves.
BTW, for the renaming requirements of downstream branches, it's not just
changing the name of the filename but the font name inside the font: the
way the font presents itself to the user via a font menu/font cache. But
in the context of webfonts linking both should probably be in sync. A
branch of oflb.org/people/benweiner/puritan-2.0-1.otf (font name
Puritan) should probably be oflb.org/people/fooauthor/foo-1.0-1 (font
name Foo) with a entry somewhere indicating the branch relationships.
I find the fontaine classification with the visual icons really
excellent: it will be a good resource.
Thanks for all the time and energy you already put into all this.
Very promising work!
Cheers,
--
Nicolas Spalinger, NRSI volunteer
Debian/Ubuntu font teams / OpenFontLibrary
http://planet.open-fonts.org
signature.asc
Description: OpenPGP digital signature
