@seantalts, is there going to be an announcement made for the release of 2.16? At some point in the past, we decided to announce the interfaces as they were ready since the processes aren’t in synch.
Question about the release notes: was @breckbaldwin added to the dev team? I don’t remember a discussion about it and I don’t see him on the team page on the website. In general, I don’t think someone ends up on the dev team automatically by being hired by Columbia (or even NumFOCUS). But either way, we should make the release notes consistent with the website and manual.
I don’t know how those two things are supposed to work, so I’m happy to take direction from you guys.
Do we need release announcements when there isn’t some critical bug? For
PyStan, I’m happy to let a new release “announce” itself via the
existence of the software on PyPI and a tag on Github.
I could be persuaded to post a little note in an “Announcements” section
on discourse though. But perhaps this could be automated too? Can we
give Jenkins a discourse account? :)
I like release announcements because we don’t do that much outbound communication to our users and it’s a good way to let people know we’re still moving. I’ve been thinking about feeding merged pull requests into twitter, but been hesitant.
We don’t need one for PyStan if the users are able to tell through PyPI and GitHub. I think the bulk of other users may need announcements.
Yup. Just something quick and possibly just the release notes in Stan would do.
@seantalts, here’s what I’d suggest:
- edit the 2.16 release notes in the Stan repo and remove Breck from the announcement.
- In the announcements section, add a post saying that Stan v2.16 is out. It can be as simple as something like this from stan-announce for v2.13.1
click for poorly formatted v2.13 release announcement
CmdStan v2.13.1 is now available! Download a tarball or zip file from GitHub: https://github.com/stan-dev/cmdstan/releases/tag/v2.13.1
Some of the noticeable features include:
- bin/stansummary can now produce a csv file
- Stan programs can now have undefined functions where the definitions are included in an external C++ header file
- Stan language:
- Vectorization of all unary functions
- Compound declare and define
- Matrix exponential is now available: matrix_exp()
For the full release notes, see:
Stan Development Team
Is that enough info? I can provide title and body of a post if you want.
I’m gonna comment on the thread saying CmdStan is out.
The easiest place to make announcements would be Twitter, I think.
If everyone is tagging releases on github it wouldn’t be too difficult to have Jenkins automatically make a tweet when releases get tagged (or maybe, 92 hours later, to give CRAN and PyPI some time).
I can look into this if it sounds like a good idea to people.
We can probably set up ifttt to automatically tweet releases.
Feel free to announce anywhere you see fit. I don’t read Twitter, but then I don’t need release announcements. Do you need credentials for the Stan Twitter account (I don’t use Twitter, so I have no idea what it is myself—I think Daniel or Michael may have it).
We should start a blog for Stan and we can release there and auto-tweet the way Andrew’s blog does.
Now that Bob mentions his non-use of Twitter (which I don’t particularly
like but which is preferable to other alternatives), perhaps just
automatically posting to Discourse would be a good place to start. It
would be nice to see “X is tagged on Github” automatically show up in
the Announcements section.
I thought the rule of thumb was to announce when the major interfaces are out (cmdstan, pystan, rstan). For 2.16 rstan is still not out…
I am not sure if users care anything about stan-math or stan being tagged.
It is out, but not mirrored, so not announced yet.
Cool. This is very fresh… this morning CRAN did still show 2.15; now its 2.16.2… great!
At some point we decided the announcements would be staggered since the process is not synchronized.
There’s no harm in letting CmdStan (and MatlabStan, StataStan, Stan.jl, MathematicaStan) users know that the interface they use has been updated.
@bgoodri waits for the CRAN mirrors to catch up before announcing. I think that’s reasonable. There’s still usually a delay between a CmdStan release and RStan, so just announcing when ready seems like the right thing to do.