It's important to note:<p>> TileMill has shifted to an open open source contributor model and moved to its own organization, tilemill-project.<p>A shift which has come a year or so after the project was abandoned by Mapbox in favor of Mapbox Studio (the official, supported successor to TileMill).<p>TileMill alone is still useful for creating one's own raster tile set, as long as the application remains in working order.<p>However, last I checked it gave me tremendous issues running on Linux as many dependencies had become outdated.<p>Although, I haven't checked back in a while since I moved to Mapbox Studio.
Check out also Kosmtik, <a href="https://github.com/kosmtik/kosmtik" rel="nofollow">https://github.com/kosmtik/kosmtik</a>, which is a similar alternative and which many in the OpenStreetMap community have shifted to using.
This would be helped by the landing page actually explaining what this is and what I would do with it. A central collection of technical links shouldn't be how you introduce a project.
I used TileMill quite a bit in 2014-15, but stopped after Mapbox abandoned it. Last I checked it wouldn't even run on MacOS anymore. The latest release was in 2012, and based on commit history there hasn't been any movement for ~8 months.<p>I was a lot of fun to work with, CartoCSS allowed me to be immediately productive with it. In the end though it was too expensive to serve tiles from AWS.
A few years ago a good share of the client browsers would choke on complex map renderings, lets say, more than a few thousand SVG complex polygons. At the time, for this kind of use case often it was faster to generate bitmap tiles at the server side. I wrote a tile server using Mapnik bindings for Python - and TileMill was of great help.<p>It is no longer the case, modern computers and browsers have acceptable performance when using vectorial maps even for very complex maps. The need for firing your own tile server is very unusual.