Issue new release/give open source community ability to create releases

[pnorman]

There have been a lot of changes since the last tagged release. We would like to create a release and update the binaries.

This means

• Tagging a new release
• Building binaries
  ** Uploading the windows build to http://tilemill.s3.amazonaws.com/latest/TileMill-v0.y.z-Setup.exe
  ** Uploading the Mac OS X build to http://tilemill.s3.amazonaws.com/latest/TileMill-0.10.1.zip
  ** Updating the PPA at developmentseed/mapbox

If I were to do this myself I'd need the access rights as well as instructions.

What is the best route for this?

[tmcw]

Conceptually, I'm into this. @springmeyer has the best and most current insight into how the release process works and any additional complexities.

Management-wise, what I'd like and what I've seen in similar projects is:

• Should we move TileMill to a different organization to simplify membership?
• Are there more people interested in a maintainer and community manager role, so that there's more of a sense of stability?
• Release-wise, can we make it clear that TileMill in the future will be community-managed instead of under the wing of Mapbox?
• Should we rename TileMill to reflect the change in status?

[pnorman]

• Should we move TileMill to a different organization to simplify membership?

Probably something to consider for the future, but I think it can happen later.

• Are there more people interested in a maintainer and community manager role, so that there's more of a sense of stability?

I sure hope so! I don't run Tilemill myself so I'm not likely to do anything beyond releases. My only interest is in updating dependencies. In short, I'm involved here because I need to be, not because I want to be. If it weren't for Windows (and partially OS X) I'd drop TM support from osm-carto and just use Kosmtik, since the additional features of TM don't work with a style sheet that complex. But TM is an easy install on Windows, and Kosmtik is not.

Release-wise, can we make it clear that TileMill in the future will be community-managed instead of under the wing of Mapbox?

This is a tougher one. I think this release should go out like other MB ones because most of the commits are from MB. In the future, perhaps we can upload the binaries as Github releases? That wasn't possible at the time of previous releases but is an option now. I'm not sure what to do with the PPA, but that's because I'm not sure what it's doing right now. I don't think anything else needs to change on the technical side.

On the branding/community side, it'd require revisions to the Tilemill content on mapbox.com. More on that with the next question

Should we rename TileMill to reflect the change in status?

I think the name should be kept. Mapbox isn't using the name for anything else and unless they want to reuse the name for a different product, there's no risk of confusion.

If Mapbox wants to make it clear that they're not actively involved in Tilemill, how about taking the TM content on Mapbox's site and splitting off tilemill.com, then redirecting mapbox.com/tilemill stuff there? This would be good for MB since TM doesn't fit with your current infrastructure or plans and by putting it on its own domain without references to Mapbox* you'd make it clear that it's no longer supported.

*Except as appropriate for mentioning Mapbox integrations, that MB wrote most of it, etc

[pnorman]

cc @nebulon42

[csytsma]

I'd love to see Tilemill kept active and more frequent releases. I'd be willing to help on release process.

[tmcw]

Two issues here:

Tagging versions

As far as I can see it, for osm-carto, that's all they really need - to be able to point to a tag rather than 'master'. This is as simple as tagging any other repo, and now that @pnorman has organizational access, he can do that.

Releasing binaries

Unfortunately, a whole 'nother story: packing desktop software is just as bad as it's cracked up to be. See tilemill-win-launcher and tilemill-builder-osx for the details there. I never ran a TileMill release, so never got the lay of the land in terms of building new distributions, and we know that they have significant gotchas - OSX startup issues and so on.

---

I think there are roughly two options:

• There are new tagged TileMill releases, but no more binary releases.
• Someone volunteers to adopt and fixup the builder process. Unfortunately the folks who wrote this process are pretty tapped with other work, so this would be basically an unsupported vision-quest.

[kocio-pl]

Thanks for your comment!

@gravitystorm said:

So I don't think that Tilemill is actually ready, but it will be after the next tag / release of tilemill.

and I understand his words like "a tag would be enough" - but I'm just guessing.

This way or another, tagging would be probably needed to make a proper binary release (not tilemill-snapshot), so I would like to see new TileMill tag anyway. With 405 commits and almost 4 years from v0.10.1, I guess v0.11.0 (instead of v0.10.2) makes a lot of sense.

[pnorman]

With the repo move + travel I haven't had a chance to respond to this

Tagging versions

As far as I can see it, for osm-carto, that's all they really need - to be able to point to a tag rather than 'master'. This is as simple as tagging any other repo, and now that @pnorman has organizational access, he can do that.

Releasing binaries

Unfortunately, a whole 'nother story: packing desktop software is just as bad as it's cracked up to be. See tilemill-win-launcher and tilemill-builder-osx for the details there. I never ran a TileMill release, so never got the lay of the land in terms of building new distributions, and we know that they have significant gotchas - OSX startup issues and so on.

Tagging a release and uploading it to npm would be a start, but we might need a release like the current TM releases. Either that or give up on binary releases completely and revise the tilemill instructions accordingly.

Put differently, what we need is the tilemill.com version to be updated. Needing this as opposed to just a git tag is because tilemill has two main advantages over alternatives like Kosmtik

• A built in text editor
• A sane way for users not familiar with the command line to install and run, particularly on Windows.

The text editor doesn't matter for osm-carto because it doesn't work on very large styles, but the binaries do.

If we can't get new binary releases going or it's too much work, what I'm inclined to do is tag a new release (easy), drop osm-carto TM support, and potentially drop the outdated binaries.

I think the route forward is to start by tagging a RC and trying to get the build process to work on that RC.

[tmcw]

Okay, you're good to go on that route, then - you have the repo permissions to tag an RC and we can deal with the s3 permissions once a binary is built.

[pnorman]

I'll look into it in a few days when I'm back home.

[kocio-pl]

I try to look at making Windows binary and it's possible, but just a lot of work. With Windows/MSIE virtual machines freely available one can do it even under Linux with VirtualBox, as I do. However big list of dependencies needed, lack of automation and time passage (some locations or versions may change) makes it rather tedious.

@springmeyer I guess you are the one who has working VM to create Windows binary version of TileMill? Could you try it with current master?

[tmcw]

You'll need to set up a VM from the mapnik instructions linked in that ticket - this project can't rely on other people's time for the build process.

[kocio-pl]

Anything new with this idea? I was trying to set Windows environment inside VM, but it proved to be really complicated and I gave up at some point, however I might try it again. People are still asking for TileMill and I guess Windows executable is important for them, as @pnorman said.

[kocio-pl]

I just learned that current Mapnik builds for Windows can be done this way:

https://github.com/mapbox/windows-builds

so I will try again.