Perhaps this is a question that has an answer elsewhere but, irrespective of if this change should be made to WMF wikis, why are we:
a) Making this a change in core?
and b) Not making the change in core be a SASS variable that can then be set as a preference somewhere? (I say this because we've consistently identified that some languages need different default fonts. If it was a preference in that we could set via our multiversion scripts it would obviate the need for overrides in common.css just to make things work.)
~Matt Walker Wikimedia Foundation Fundraising Technology Team
On Tue, Apr 8, 2014 at 2:03 PM, Martijn Hoekstra martijnhoekstra@gmail.comwrote:
On Tue, Apr 8, 2014 at 10:20 PM, Steven Walling <steven.walling@gmail.com
wrote:
On Tue, Apr 8, 2014 at 12:33 PM, Martijn Hoekstra < martijnhoekstra@gmail.com
wrote:
On Tue, Apr 8, 2014 at 8:13 PM, Erik Moeller erik@wikimedia.org
wrote:
On Tue, Apr 8, 2014 at 10:59 AM, Martijn Hoekstra martijnhoekstra@gmail.com wrote:
So, the font stack changes with regards to the status quo now
change
nothing for Windows users, changes Helvetica -> Helvetica neue for
Mac
users and changes Arial, DejaVu Sans or Arimo for possibly
something
else,
amongst which Nimbus Sans L, maybe, maybe not.
Actually, it's a bit more complicated. All users get serif fonts for headings, which they didn't before and which is probably the biggest visual before/after difference. The serif fonts still prioritize free/libre fonts over non-free ones.
The body fonts prioritized free/libre fonts on deployments, but for Windows users without ClearType/anti-aliasing, those render very poorly, so they were disabled shortly after deployment. This is now causing people to be upset because the initial agreement to never prioritize non-free fonts is no longer maintained for the body.
Odder's patch would revert to sans-serif as a generic classification for the body, but doesn't touch the font specification for the
headers
(yet). The commit summary is a bit misleading in that regard.
Yes, I should have made that clear: I do very much support the Odder patch[1] ( https://gerrit.wikimedia.org/r/#/c/124475/ ) that reverts
body
to sans serif and keeps @content-heading-font-family: "Linux
Libertine",
Georgia, Times, serif;
That is not the status quo, but the diff between the Odder patch and
the
typography refresh basically is the "Set a non-free font stack to give
Mac
now Helvetica Neue rather than Helvetica", with a -2 is planted in the ground before as a demarcation line. That's the point that I don't
think
is
worth having a non-free font-stack for, and that I certainly think
standing
your ground for the brave new world of typography refresh is
constructive
for.
This is a persistent myth floating around about this. Neue for Mac users
is
most definitely not the only effect of explicitly declaring the stack. As Jared, S Page, and others have already pointed out, and as is stated in
the
FAQ on mediawiki.org, the impact of the current stack is:
-- Linux users get Nimbus Sans L, instead of DejaVu Sans, Liberation
Sans,
or whatever else the default sans is on your distro. Nimbus has an
improved
x-height and is much more consistent with the other sans-serifs we're specifying.
Unfortunately, we didn't properly test this. With the large diversity in results of the tests we did do on https://www.mediawiki.org/wiki/Typography_refresh/Font_choice/Test I doubt it's a hard rule that Linux users will now universally get Nimbus Sans L. I'm also not sure that Nimbus Sans L is the superior alternative over Liberation Sans. I'm by no means an expert, but in the tests on https://www.mediawiki.org/wiki/Typography_refresh/Font_choice it scored exactly the same as Helvetica Neue (both in the good an the bad). Nimbis Sans L unfortunately hasn't been part of that test.
-- Windows users always get Arial, unless they have Helvetica installed.
This means many of the Windows users who might otherwise have set an alternative sans in their browser default (like Verdana or Tahoma) will
now
always get Arial.
Windows Helvetica seems to be generally a lie, and is not Helvetica but "Helvetica" (actually Arial). I don't see overriding a users preference with our preference as an advantage. Most people don't change their default font in their browser. In those cases it's probably good if we work with our preferences instead. If someone put in the effort to set a different preferred font probably did that for a reason that we shouldn't override. I assumed that we only wanted to override "unset" user preferences, and not actually override the settings that users have explicitly chosen. I was apparently mistaken in that.
-- And yes, Mac users or those with it installed get Neue Helvetica instead
of older version. This is a minor but worthwhile improvement for Mac
users.
For example, Neue Helvetica actually has properly designed font weights
for
bold, italic, etc. so that the cap height and x-height are consistent between weights. Regular Helvetica was really not consistently designed across weights.
Declaring some of the system defaults explicitly is not only an
improvement
for Mac users. It means that regardless of whether you switch between devices/browsers, you always get a grotesque/neo-grotesque sans-serif ( https://en.wikipedia.org/wiki/Vox-ATypI_classification) which is ideal
for
reading long, large blocks of text and is more neutral.
[1]My only nitpick about it is that I'm wondering what Times is doing
in
that stack. I can't think of any situation where a user wouldn't have
Linux
Libertine or Georgia, but does have Times, yet doesn't have it as its default serif font. When one has specifically set a default serif
different
from Times, you probably have a good reason for it - or at least a
better
reason than the websites desire for Times, and we should respect that.
Yet
this beef is very small compared to all other issues in this thread.
Times, like setting Helvetica, is there because it's what Linux systems recognize and match to. Linux fc-match has no idea what Georgia and Linux Libertine are unless you've installed them. By setting Times, we ensure that Linux uses Nimbus Roman No9 L for headings, which complements the
body
typeface selected on Linux well and which is consistent in style with the other serifs specified.
This comes down to the same thing I said under Windows. I was under the impression that not overriding the explicit choice of font a user has made is a good thing. That impression is apparently up for debate. But this isn't really a big deal to me.
A lot of this stuff is already documented in the FAQ on [[Typography refresh]] on mediawiki.org. We produced it to answer the basic questions just like this.
With the free fonts removed from the stack, unfortunately the FAQ doesn't
provide all the answers anymore. That's no criticism on the FAQ, after the quick turnover of events it's not surprising that it's not up to date anymore.
There's some additional discussion about Georgia as a font choice due to its use of text figures (AKA old-style numerals), which some
people
find look odd in headings with numbers, especially in non-Latin scripts where old-style numerals may not be commonly encountered. Due to this, some are arguing for also changing the style for headings to serif (_not_ sans-serif) as a generic classification, or removing Georgia from the stack. That particular issue hasn't been discussed
in
detail yet, as far as I can see.
I think the differences of opinion here are not worth a holy war. Prioritizing a non-free font before free ones for the _body_ with a clear FIXME indicating that this is not a desirable state is IMO only marginally different from reverting to sans-serif until we have a free/libre font that _can_ be prioritized for the body. So I think either outcome should be OK for the short term, and we should focus
on
the longer term question of a good font stack for the body that prioritizes free/libre fonts.
Let's not polarize each other too much. All the arguments I've heard have been fundamentally reasonable and rational, not just "Change is evil". Some people hate the serifs per se, but that's a smaller discussion compared to these conversations, which are about substantial things that can be reasoned about.
Erik
Erik Möller VP of Engineering and Product Development, Wikimedia Foundation
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l