On Sun, Apr 6, 2014 at 4:05 PM, Tomasz W. Kozlowski tomasz@twkozlowski.netwrote:
1. I am deeply uncomfortable with the fact that you are choosing un-free
fonts over free ones. 2. I am deeply uncomfortable with the fact that you decided not to respect the consensus /not/ to choose non-free fonts -- such as Arial and Helvetica -- over free fonts; a discussion which I only read, but which, as far as I remember, saw participation from yourself, Quim, Greg, and some other people.
We've tried the alternative and it's untenable according to the feedback we're getting. I wish it wasn't. I'd rather put free fonts first in the stack, if they actually work for users. Twice now we've tried putting different freely-licensed fonts first. Both times, Windows users who had them have told us they either merely disliked them or they have caused unacceptably poor rendering, particularly for those without font smoothing. There simply is not widely-available font that meets all our needs while also being freely-licensed. The compromise is either to deliver a freely-licensed webfont to all users (which we're not going to do right now, though it's the ideal IMO) or to specify the best fonts users already have on their system free or not, which accomplish the consistency and legibility we're looking for. This is just the reality. Whether or not the CSS/LESS declares them explicitly or not, non-free fonts are what most users have already and want to use, because they actually work. This is true whether we set a more specific stack than "sans-serif" or not.
As for your suggestion that I'm only looking to make a fuss, here's some basic facts for you to ponder.
A. /I/ pointed it out to Greg and to you on IRC that deploying Typography Refresh to all wikis on the same day (March 28) was a bad idea, and that it would be better to roll it out with MediaWiki 1.23wmf21, as it would give time to inform the community (as well as to push some last-minute fixes).
Delaying release to anticipate bugs that have not yet been reported by anyone makes no sense. At the time of release there were only four bugs open related to VectorBeta as an extension, none of which could have told us about the issue. How could last minute fixes be pushed for a bug no one had actually reported yet?