Closed pnorman closed 10 years ago
Hmm, I'm torn on this one. If you were to ask if we should bundle natural earth data with the repo I'd say no, since they are large read-only datasets that I wouldn't want in the repo. My thoughts are similar for the fonts, since we're just bundling assets rather than them being something that we want to maintain.
On the other hand, sourcing all the fonts can be a pain for non-Ubuntu people. Does anyone have any feedback on this point e.g. from Mac or Windows users who have installed these stylesheets?
But of course, if a particular font is missing that shouldn't be a great problem for many users/testers/developers of the style. It's non-fatal to working with the style in Tilemill.
On the other hand, sourcing all the fonts can be a pain for non-Ubuntu people. Does anyone have any feedback on this point e.g. from Mac or Windows users who have installed these stylesheets?
If you're not concerned with packaging systems, it's not too hard to install the fonts on Windows. The problem I've seen is more on Linux where different font versions are packaged on different distributions.
I'm really wondering about FontConfig and virtual fonts, but first I need to try them out.
dpkg -L ttf-dejavu fonts-droid ttf-unifont fonts-sipa-arundina fonts-sil-padauk fonts-khmeros ttf-indic-fonts-core ttf-tamil-fonts ttf-kannada-fonts | grep '\.ttf' | xargs du -hsc {}
tells me all the TTF files from the packages add up to 30MB, of which half is unifont. That's probably a bit much, unless virtual fonts turn out to be a significant advantage.
We can use a relative font directory to include fonts relative to the stylesheet. We've also had bugs come up with font packages changing names (#837) and different font versions having different names (#429).
Is it worth considering packaging fonts with openstreetmap-carto. This would
Cross-ref https://github.com/gravitystorm/openstreetmap-carto/issues/632#issuecomment-46743487