Dear all,
My apologies up front for the long e-mail that follows. In this e-mail you
will find a comprehensive status overview of the recent WebFonts deployment.
On Monday December 12 at 18:00 UTC we deployed the extension WebFonts[1] to
40 wikis in 11 Indic languages and Wikimedia Incubator -- all wikis in
Assamese, Bengali, Gujarati, Hindi, Kannada, Marathi, Nepali, Oriya,
(Eastern) Punjabi, Sankrit and Telugu have WebFonts now. WebFonts was not
deployed on Malayalam and Tamil projects. The reason for this was that
community members had requested us not to. We are confident that in time,
the communities will request that WebFonts is enabled on their projects.
WebFonts aims to resolve the issue that users see incomplete web pages,
because the fonts to properly render the page is not present in the local
system by downloading the font through the browser.
One of our great challenges developing this functionality is the multitude
of scripts and the low availability of freely licensed fonts that may be
modified and redistributed.
Over the past few months we have tried to build out a collection of fonts
in the extension mainly for Indic languages, and we have performed many
tests. We have solicited community involvement through messaging in village
pumps, e-mails on mailing lists, blog posts on personal blogs as well as on
the Wikimedia Foundation blog, at developer events, through personal
e-mails and through our bug tracker, and gotten some feedback, although
unfortunately not for all the languages we would like to have gotten it
for. We will of course continue our efforts in this area. Next to the
community involvement, we have had a two day session with the Red Hat
Localisation team in Pune, India.
Since the deployment, we have been criticised for not communicating enough
-- or not through the right channels, not with the right people, not in
time, or too soon, or not with the right messages. I'm not really sure how
to respond to that, except for uttering a general "mea culpa, mea maxima
culpa". We are working really hard in continuously improving the work that
we do, and the way that we do it. We make mistakes, we are human after all,
and when we become aware of our mistakes, we will do everything in our
power to make it better.
With our team we support the mission of the Wikimedia Foundation to
"imagine a world in which every single human being can freely share in the
sum of all knowledge." I care about that -- a lot. We all care, and I am
pretty certain that we're not ignorant, dismissive or incapable. I
acknowledge that we as the Localisation team are a relatively new entity
within the MediaWiki development community and within the Wikimedia
Foundation, with a very wide scope, and that we are dealing with a lot of
technical details on which we are simply not able to assess the final
quality; there are after all 7.500 languages in this world of over 7
billion people that we theoretically all cover, some 350 of those languages
are supported in MediaWiki, and 280 within Wikimedia.
I accept that we cannot keep everybody happy -- doesn't keep us from
trying, though. I want to try and work with as many people as possible in a
constructive way. With these numbers, that's not always easy to coordinate.
To channel the input on languages, we have set up "Language Support
Teams"[2]. We do not yet have a language support team for every language.
Please sign up if you care about the technical facilitation of your
language in the Wikimedia movement. Let's use the mediawiki-i18n mailing
list[3] to have constructive discussions about language support. Let's use
the #mediawiki-i18n IRC channel[4] on Freenode to have real-time
discussions. Let's use bugzilla.wikimedia.org to report bugs[5]. Link [5]
explains the bug reporting procedure. If you already know how, report
issues quickly using this link: http://ur1.ca/6ov9a .
Since the deployment, we have been made aware of about 17 issues. Some very
serious in nature, others not requiring immediate attention. Yesterday an
issue with web fonts not loading in Firefox was resolved in the
infrastructure. Today around 15:30 UTC, we have deployed fixes for an
additional hand full of issues[6]: functionality disabled in IE6, IE8 on
Windows XP, selection buttons not working properly in IE7 and hiding the
Samyak fonts in the font selector. During our current sprint, we are
working on a framework for multi-lingual and localised user documentation
as well as feature based feedback functionality for WebFonts, Narayam and
Translate. In the future we will also explore what is known as "dark
launch" by some, a kind of hidden live deployment of a feature, only usable
be for example manipulating a URL. This would allow us to deploy a feature
in a live environment, without having the "full deployment" impact.
Thanks for reading through this. I am looking forward to working with you!
Please read on for details on all the issues that were reported on WebFonts
recently.
Cheers!
Siebrand Mazeland
Product Manager Localisation
Wikimedia Foundation
=======================================
Links
=======================================
[1] https://www.mediawiki.org/wiki/Extension:WebFonts
[2] https://translatewiki.net/wiki/Language_support_team
[3] https://lists.wikimedia.org/mailman/listinfo/mediawiki-i18n
[4] https://translatewiki.net/wiki/Special:WebChat
[5] https://www.mediawiki.org/wiki/Bugzilla
[6] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106204
=======================================
Open issues
=======================================
https://bugzilla.wikimedia.org/33004 -- Old cached pages do not have web
fonts enabled
Priority: HIGH
--------------------------------------------------------------------------------------
Wikimedia is able to serve this many pages with relative few servers
because of very aggressive caching strategies, especially for anonymous
users. WebFonts requires the addition of JavaScript for anonymous users,
which is not being done for pages that are in the squid cache at the moment
WebFonts was enabled. All squid cache objects for wikis on which WebFonts
was deployed need to be purged. An internal RT ticket created for the
Wikimedia Operations team to get anonymous squid caches purged. This may
take up to a week or longer to be resolved.
https://bugzilla.wikimedia.org/33018 -- Firefox 5 on Windows XP has script
time-outs
Priority: MEDIUM
--------------------------------------------------------------------------------------
The Localisation team has tested this report, and was not yet able to
confirm the observation. The reason for using a non-recent version of
Firefox for the report was the alleged lower memory usage. Brion noted that
Mozilla has been actively working on lowering memory usage over the last
year, so the reporter may be better off with the current versions than the
old ones.
https://bugzilla.wikimedia.org/33110 -- Google Crome on Windows XP dispays
gibberish
Priority: LOW
--------------------------------------------------------------------------------------
Observed very rarely on a page on Wikimedia Incubator, and we have not been
able to reproduce this observation, let alone reproduce it reliably. A
screenshot is present in the bug report. Except for reporting upstream, no
action is being taken on this issue at this point in time.
https://bugzilla.wikimedia.org/33054 -- Hinting issues in Lohit fonts
Priority: MEDIUM
--------------------------------------------------------------------------------------
Confirmed in Windows XP. We can do something to the font by adding hinting,
but this is a lot of work if it needs to be done manually. The stem of the
Lohit glyphs could do with more width and darkness. This may not be
desirable for platforms (Linux) which render it perfectly, because it
already has hinting and anti-aliasing on an operating system level. Same
goes got Windows 7.
https://bugzilla.wikimedia.org/33100 -- Page crashes on Webkit browsers
with WebFonts enabled.
Priority: MEDIUM (could be HIGH if we find many occurances)
--------------------------------------------------------------------------------------
A page in Nepali Wikipedia makes a tab on Mac OS X 10.7.2 with Google Crome
crash. This behaviour was also reported for Mac OS X 10.7.2 (11C74) with
Safari 5.1.1 (7534.51.22, r102522) [This is a webkit nightly build] by
thedj. This is most probably related to the WebFonts code, because if, as a
logged in user, web fonts is disabled in preferences, the page does not
crash Chrome.
Developer Derk-Jan Hartman was asked to report this bug in the WebKit.
Please make us aware of any additional pages that would cause this
behaviour in any wiki.
https://bugzilla.wikimedia.org/33102 -- OSX 10.7.2/Opera 11.60 has no
fallback for Latin characters
Priority: MEDIUM
--------------------------------------------------------------------------------------
This is a bug that needs to be reported upstream. No technical measures
have been taken so far to mitigate this issue. One of the Localisation team
members has been in contact with a high level executive of Opera, and will
contact that person again. We're going to wait for a few days for an
outcome -- if there is no expectation of a relatively quick fix, we might
disable WebFonts for Opera completely. Opera unfortunately does not have a
public bug tracker.
https://bugzilla.wikimedia.org/33027 -- Narayam and WebFonts both loading
slows down page
Priority: MEDIUM
--------------------------------------------------------------------------------------
The reporter claims that the functionality is quicker on
translatewiki.netthan it is in Wikimedia wikis. A commenter states
that more functionality
usually means more code, means more data that needs to be transferred, and
without changing bandwidth, that causes longer load times.
This currently isn't our highest priority, but eventually we will look into
this a little deeper. We're inviting volunteers to do some of the data
gathering and analysis for us. What is needed in our opinion is insight in
the data volume added by WebFonts, as well as an assessment of the code
quality with regards to size optimisation. All referenced properly, of
course :). There are alternate EOT conversion tools that have a good
compression ratio. Needs to be explored, but EOT is not required for modern
browsers since they started using WOFF fonts which are compressed OpenType
fonts.
https://bugzilla.wikimedia.org/33085 -- Integration of updated Lohit-Tamil
Font
Priority: MEDIUM
--------------------------------------------------------------------------------------
Request to update WebFonts with a font that is updated upstream. This is
something the Localisation team checks regularly. Will probably be closed
this week, pending issues the have a higher priority.
https://bugzilla.wikimedia.org/32942 -- Provide help page and bug report
link for WebFonts
Priority: HIGH
--------------------------------------------------------------------------------------
More recently developed tools by the Wikimedia Foundation have often
included feedback mechanisms. The Localisation team plans on implementing
these for the functionality of the WebFonts, Narayam and Translate
extension. Besides that, we also want to provide multi-lingual and
localised documentation. This needs some thinking and some work to provide
in a structured and navigable way. We'll keep you posted. It will most
probably involve translatable *user* documentation on MediaWiki.org and
hopefully it is possible to have one feedback location per feature across
the multiple Wikimedia wikis -- this is something we're going to contact
the ArticleFeedback and MoodBar teams for.
=======================================
Closed issues
=======================================
https://bugzilla.wikimedia.org/33025 -- When changing to a non-default web
font, the content does not
--------------------------------------------------------------------------------------
This issue was a side effect of a feature to allow multiple web fonts to be
used using the "lang" attribute. It was resolved in
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/105980 and has been
deployed.
https://bugzilla.wikimedia.org/33034 -- Web fonts not loading in Firefox
--------------------------------------------------------------------------------------
Duplicate reports were 33038 and 33044. This issue originated from
http://www.w3.org/TR/css3-fonts/#same-origin-restriction. Almost all
browsers except for Firefox ignore that specification. A fix was designed
and deployed: https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106092,
https://gerrit.wikimedia.org/r/1501. Thanks to Roan, Brion and Ryan for
their help.
https://bugzilla.wikimedia.org/32775 -- Gibberish in Internet Explorer 8 on
Windows XP
--------------------------------------------------------------------------------------
This is an unexplained phenomenon only observed in Internet Explorer on
Windows XP. It is also hard to reproduce. One of the developers was able to
make something somewhat reproducible on a clean, fully patched installation
of Windows XP with Internet Explorer 8. See bug report for details.
Based on these observations we think it is a bad idea to keep supporting
WebFonts in Internet Explorer 8 on Windows XP and we have disabled it in
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106172. This fix has
been deployed.
https://bugzilla.wikimedia.org/33096 -- Internet Explorer 6 does not have
font fallback
--------------------------------------------------------------------------------------
IE6 not having font fallback causes Latin characters to display as squares
when a web font is loaded that does not contain glyphs for the Latin
script. A screenshot is available at
http://media.crossbrowsertesting.com/users/34057/screenshots/window/z669002….
Based on this observation, we think it is a bad idea to keep supporting
WebFonts in Internet Explorer 6 and we have disabled it in
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106172. This fix has
been deployed.
https://bugzilla.wikimedia.org/33024 -- WebFonts menu buttons not working
in IE7
--------------------------------------------------------------------------------------
This was caused by the JavaScript $( '<input type="radio" />' ) . attr(
"name" ,"font"); not working in IE6 and IE7. Updating name attributes once
they have been created is not possible. We think there may be more
occurances of this in our code (one occurance in jQuery has already been
identified: resources/jquery/jquery.validate.js:59). A fix was made in
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106175. This fix has
been deployed.
https://bugzilla.wikimedia.org/33040 -- Overlap in Samyak font for Hindi
and Sanskrit
--------------------------------------------------------------------------------------
This issue occurs in Windows XP and Windows 7 (possibly also in Windows
Vista) when using Google Chrome. It is not observed when using Chrome with
Mac OS X 10.7.2 or several Linux distributions (Debian and Fedora). Samyak
Devanagari is available as a non-default web font in Hindi, Marathi, and
Sanskrit. Samyak Gujarati is available for Gujarati as a non-default font.
This font needs to be corrected. The maintainers will be notified of the
observed issues, and mean while, the fonts will be removed from the
WebFonts selection list (but can still be used using the font-family
property. A fix was made in
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106179. This fix has
been deployed.
https://bugzilla.wikimedia.org/33039 -- Overlap in Madan font for Nepali
--------------------------------------------------------------------------------------
This report was invalid. The reporter was not aware of the correct glyph
for the Nepali script.
Comments on this bug report resulted in two odd observations (Crome crash,
Opera font fallback), that have been split off into separate bug reports:
https://bugzilla.wikimedia.org/33100 and
https://bugzilla.wikimedia.org/33102.
https://bugzilla.wikimedia.org/33095 -- WebFonts menu can expand off the
screen
--------------------------------------------------------------------------------------
If the translations for "Select font" and "Login / Register" are really
short, like in http://mr.wiktionary.org, expanding the WebFonts menu for
anonymous users will display a menu that is partially off the screen. It
was resolved in http://www.mediawiki.org/wiki/Special:Code/MediaWiki/106186,
http://www.mediawiki.org/wiki/Special:Code/MediaWiki/106197,
http://www.mediawiki.org/wiki/Special:Code/MediaWiki/106201,
http://www.mediawiki.org/wiki/Special:Code/MediaWiki/106202. These
revisions also depend on a few small UI changes of both WebFonts and
Narayam, and will be deployed on December 19, 2011.
<no bugzilla report> -- WebFonts menu expands under the control for
customised input method in IE6 on transliteration
--------------------------------------------------------------------------------------
There are issues with the z index in IE6. Because of
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/106172, WebFonts is
no longer available in IE6, so this issue is obsolete. Observing that the
Hindi projects Wikipedia and Wiktionary are using an custom input methods
tool, we would like to invite them to test Narayam which contains many
input methods in a MediaWiki extension. We are very open to having the
Hindi input method InScript tested and add a transliteration input method
with some community representatives, as we have done with other Indic
languages. We hope this will eventually lead to Narayam being adopted by
the Hindi community, and the custom input method being abandoned.
What would be the right way to differenciate between fallbacks like
Picard -> French (there's a similarity between the languages, the
message might even be the same) and French -> English (which is there
just for having a base language). Take the fallback lists like shown in
the message file, I guess? What's the prefered way to do that? Shouldn't
we add in that case another parameter to Language::getFallbacksFor() to
disable the addition of en as final fallback?
Regards
FYI
Gerard
Sent to you by GerardM via Google Reader: CLDR Version 22 Released via
The Unicode Blog by Unicode, Inc. on 9/10/12
Mountain View, CA, Sept. 10, 2012 - The Unicode® Consortium announced
today the release of a new version of the Unicode Common Locale Data
Repository (Unicode CLDR 22.0), providing key building blocks for
software to support the world's languages.
Unicode CLDR 22.0 contains data for 215 languages and 227
territories—654 locales in all. The main focus for this release is to
flesh out data items in major languages and locales, yielding an
increase of over 100% in the total number of data fields. Other major
features include the addition of keyboard mapping data for different
platforms, the new Zhuyin (Bopomofo) sort order for Chinese, and script
metadata. There are also enhancements to compact decimals (such as
formatting 1,000,000 as “1 million” or “1M”) for different languages
and to rule-based number formats (such as writing 423 as "four hundred
and twenty-three"). For more details, see the CLDR 22.0 Release Note.
CLDR is used to adapt software to the conventions of different
languages for such common software tasks as formatting of dates, times,
time zones, numbers, and currency values; sorting text; choosing
languages or countries by name; and transliterating different
alphabets.
It is by far the largest and most extensive standard repository of
locale data, used by a wide spectrum of companies for their software
internationalization and localization. It is widely deployed via
International Components for Unicode (ICU), and also accessed directly
by companies such as Apple, Google, IBM, Twitter, and many others.
CLDR is part of the Unicode locale data project, together with the
Unicode Locale Data Markup Language (LDML)—an XML format used for
general interchange of locale data, such as in Microsoft's .NET. See
the charts pages for views of the CLDR data, organized in various ways.
For more information about the Unicode CLDR project see
cldr.unicode.org. About the Unicode Consortium
The Unicode Consortium is a non-profit organization founded to develop,
extend and promote use of the Unicode Standard and related
globalization standards. The membership of the consortium represents a
broad spectrum of corporations and organizations in the computer and
information processing industry. Members are: Adobe Systems, Apple,
Google, Government of Andhra Pradesh, Government of Bangladesh,
Government of India, IBM, Microsoft, Monotype Imaging, Oracle, SAP,
Tamil Virtual University, The Society for Natural Language Technology
Research, The University of California (Berkeley), Yahoo!, plus well
over a hundred Associate, Liaison, and Individual members. For more
information, please contact the Unicode Consortium
http://www.unicode.org/contacts.html.
Things you can do from here:
- Subscribe to The Unicode Blog using Google Reader
- Get started using Google Reader to easily keep up with all your
favorite sites
