I’ve been posting announcements of new LG versions in the Community/Announcements forum for a while now, but I’ve received a comment from someone that it might not be appropriate to post things like patch release announcements there. (Apparently people are email-subscribed and posts like these are unappreciated.)
Should I refrain from posting when we issue bugfix releases that folks might want to know about? What’s a reasonable threshold?
I think you could follow semver and do a new ANN topic when there’s a new major version, and post minor version bumps in the thread of the current major version? That’s how DataFrames does it I think (although tbf they’re not on 1.0 yet so I don’t know whether we’ll get a new topic for DataFrames 1.0…)
is a little bit over the threshold.
To decide on your own I would suggest to extrapolate what would happen if all packages would do it like this.
But on the other hand, it’s not that all other packages do this, so it’s just a minor issue here.
I would say Minor release is an appropritate threshold.
If all you are telling me is that bugs are fixed and there is no new features, or any API changes, then what am I going to do with this information?
Is it a reminder to run ]up?
I don’t feel a huge need to be reminded to do that personally, I do that pretty frequently as is.
I guess though others might be more discourse centric and want all there news there.
But i that case maybe it would reach the right people fastest by posting on existing thread.
Major also would be reasonable,
I guess it depends how much discussion there is on each release.
If your minor releases have comment threads dozens of posts long, it would get annoying to navigate if they were all in 1 post.
I personally go even further and do post ever.
Either at initial release, initial creation or 1.0.
But thats more because I am lazy and don’t tend to post much on discourse.
I don’t think its a big deal what you do, but I wouldn’t post a new thread for a patch.
To be fair i wouldn’t even tell people I made a patch release.
I often release 2 or even 3 patch releases in a day, because I follow the Continous Delivery practice of making a release after every nonbreaking PR.
So if someone makes a PR in the morning and I merge it I will release, and if someone makes on in the afternoon then that will be another release.
So I would be posting release announcements nonstop
this is a good point. Yes, my reasoning was to say, “hey, we fixed this bug that’s existed since the beginning of LG, and if you were relying on this function to produce correct results, well, now you should update and rerun your code”. The fact that we’ve had two such bugs reported and fixed in the past month is both disconcerting and welcome.
(That is to say, the two most recent bugs were errors in CORRECTNESS, which are most insidious, because valid output is provided – it’s just wrong.)
The larger issue for me is that I don’t know who’s using the code, where, and for what purposes, and when we fix a bug like this it could have some impacts I’m unaware of. Perhaps it’s the users’ responsibility to monitor the github repo, but for bugfixes I thought it made sense to post an announcement.
The general sentiment I’m hearing, though, is “nope, too much, don’t do this”, so I’ll refrain from posting bugfixes here (or will try @nilshg’s interesting approach once we hit v2.0.0).
If you’re using LightGraphs, please monitor the repo
I think correctness bugs are a big deal and am happy to see attention drawn to them-- after all, if I had used that code to make a decision etc and then I learn there was a correctness bug, maybe I should re-validate that. If it’s not code you are running regularly then ] up isn’t too relevant, but knowledge that there was such a bug is.
FWIW there was a correctness bug in Convex.jl that I ran into in the summer 2018 when I just started using Julia and I ended up switching projects because I didn’t understand the results I was seeing (and had other things to work on instead). I tried using Convex again a year later, knowing more Julia, and was able to track it down. I think if in the meantime someone had found the bug and announced it, it would’ve made things easier for me. (And maybe I should’ve made an announcement? At that point it was like my 2nd ever open source contribution, so I didn’t really think of doing anything like that).
I think you should post if you have something to tell. That might be big breaking changes that users should be aware of, exciting new features, but I think (important) bug fixes are fine as well. It might help the reception of your announcement post if you frame it in a way people understand why you go through the trouble of creating a post for a bugfix.
Ideally people who are affected by a particular bug should just subscribe to notifications on Github etc.
I think that it is best to use Announcements for initial releases and major changes. The intended audience should be people who are not already users of the package, and may want to try it out. For everything else, blog posts and release notes are better. (Generally, a significant fraction of announcements should be in blogs IMO).
That said, I agree with @oheil that it’s not a big deal at the moment since very few package maintainers make frequent announcements.
Yes, especially since one can opt for subscription to new releases only. This is hard to do for a new package one is not aware of, so announcements of those are of course valuable, but for small releases the mechanism to subscribe to releases is already there and works well.
True, but it would be not be sustainable if everyone announced every new release here. If one is interested in that level of updates one could simply watch General. There were about 380 new releases the last week alone