New candidate JEP: 369: Migrate to GitHub
Erik Helin
erik.helin at oracle.com
Fri Nov 15 10:06:05 UTC 2019
Hey Florian,
thanks for chiming in!
On 11/14/19 8:06 PM, Florian Weimer wrote:
> * Erik Helin:
>
>> There is no such thing as "switching away from mailing lists to GitHub".
>> All activity on GitHub is mirrored on the OpenJDK mailing lists.
>
> Do you plan to use comments as a way to interact with various bots?
Yes.
> Will those be mirrored as well?
No.
> It's hard to pin down what I find so off-putting in the automated
> Github mail output. I suspect that one of the issues is that you
> often have a thread which contains solely of directive to bots and
> terse one-line statements. Sure, the data is all there, but it's
> either not very relevant to me as a reader (why would I care that
> someone has scheduled a trybot run?) or not very pleasant to read.
No, I agree that would not be very pleasant to read, which is why even
the initial version of the mailing list bridge did not mirror comments
consisting of commands :)
Personally I'm using the skara-dev mailing list as the primary way to
keep up with project Skara activity so its very important to me that we
have a good mailing list experience. The mailing list bridge is *not*
just a simple "send all comments created as a separate email". For the
initial "RFR" email we decorate the email with useful information for
reviewers, see for example [0]. Everything below the "----------------"
lines is added by Skara. We _automatically_ provide reviewers with:
- commits in the PR and their subjects
- link to (automatically generated) webrev
- link to JBS issue
- direct link to the patch
- direct link to the "Files Changed" tab in the PR
- statistics for the patch
- git fetch command
- link to the pull request
All the above is added to ensure that reviewers always have access to
the information they need. We are of course more than happy if anyone
has feedback on any additional information they would like to see here.
For comments and replies we really strive to construct emails suitable
for a mailing list. For example if you review on GitHub and add three
review comments, then we don't send three separate emails (a mailing
list user would never have done that). Instead we coalesce the three
comments into a single email and also quote the relevant source code
lines being commented upon. If the user provides a general review
comment (in the "Review changes" dropdown) then that will be included
prior to line-based comments. See [1] for an example.
(Please note that there are still a few bugs to work out with replies
and the mailing list bridge, Robin is currently working on that).
I have been active on the OpenJDK mailing lists for 7+ years and have
participated on other mailing lists long before then. I hope that I can
bring some of this experience into Skara's mailing list bridge have a
feeling for what the result should look like :)
> With email, people make more of an effort to produce something useful
> to their (email) audience. Practices which have become appropriate
> for web tools (like treating comments as some sort of command line) do
> not translate to an email audience, unfortunately.
To be fair I think we see all kinds of emails on the OpenJDK mailing
list with sometimes very different formatting/content conventions. There
are those that top-post, those that aggressively trim quotations, those
that always start with a greeting, those that always end with a
"Thanks", "Cheers" or similar, those that use signatures, those that use
localized quote headers, those that do not even use quote headers, etc.
Will this change for better or worse with comments on GitHub? I think it
comes down to the precedent we as a community set. I've seen very terse
emails on OpenJDK mailing lists and I've read very enjoyable comments on
GitHub (and vice-versa of course).
Thanks,
Erik
[0]:
https://mail.openjdk.java.net/pipermail/skara-dev/2019-November/001015.html
[1]:
https://mail.openjdk.java.net/pipermail/skara-dev/2019-November/001056.html
More information about the discuss
mailing list