22 Nov
2016
22 Nov
'16
1:16 p.m.
On 22 Nov 2016, at 12:13, Guillaume Delhumeau
<guillaume.delhumeau(a)xwiki.com> wrote:
Hi everybody.
Today I would like to speak about an issue that annoys me for years.
We are working on a tool whose one of the objectives is to stop scattering
information in multiple places. It's even the main argument explained in
the video integrated on the home page of XWiki:
https://www.youtube.com/watch?v=9QTWrZ7OfzI.
But on the other hand, we, developers of XWiki, do the opposite in
practice. We discuss on mailing lists that are archived on Markmail, we
report issues on Jira and we do investigations on design.xwiki.org, and I
don't even count Github.
Honestly I don’t see the relationship between the tool we develop and how it’s developed.
These are completely separate things!
Using a car is pretty easy but creating one is hard. That’s normal and expected :)
Now if your point is to brainstorm about how we could improve the tools we use to develop,
sure, let’s brainstorm! :)
I like that we have several tool such as:
- one tool for the issue tracker (jira)
- one tool for proposal/decisions (mailing list and nabble forum - bidirectional)
- one tool for archiving discussions (markmail)
- one tool for design proposals (wiki)
- one tool for storing code (github)
- one tool for translations (l10n)
- one tool for live discussions (IRC)
- one tool for documentation (xwiki.org)
- one tool for quality reports (sonar.xwiki.org)
- one tool for CI (jenkins at ci.xwiki.org)
- one tool for storing our artifacts (nexus, at nexus.xwiki.org)
- one tool for storing xwiki extensions (extensions.xwiki.org)
Is that complex? Yes. Is development easy? No. Do we want to do it well? Yes.
They’re all required! All the tools are listed here BTW:
http://dev.xwiki.org/xwiki/bin/view/Community/DevelopmentPractices#HGeneral…
:)
Now we can discuss and decide if each of these tools is the best. For example replacing
some ML with a forum other than Nabble; don’t forget that we do have a forum at
http://dev.xwiki.org/xwiki/bin/view/Community/Forum. The main problem I see with Nabble is
that it requires 2 sign ups.
A newcomer have to understand the role and the
functioning of each tool. It's quite complex.
I don't think we should give up Jira because it is the best tool available
in its domain and there is no serious competitor.
However, concerning the mailing lists, it's very different. Let me list
some problems:
- We recently had troubles with some emails that were lost because of
subtleties in the email protocols.
- Someone who just want to discuss once have to register to the mailing
list and then receive thousands of emails every year.
- Some emails are lost in the SPAM catchers.
- You cannot use serious text formatting. As far as I know, HTML is not
supported on the ML nor in Markmail.
- You cannot send attachments.
- People looking at messages on Markmail do not always understand how to
answer (I've seen some people trying to answer directly on Markmail because
they believe it WAS the messaging tool).
- This is "so 90s"!
It still have some advantages:
- Users can use their beloved email client.
- Mail lists are quite standards in the Open Source world.
However it does not balance the drawbacks.
To fix this, I see several options:
Also see these discussion threads from the past:
* "Do we need a forum?: http://markmail.org/message/gbdnyb7jbh4ha5ja
* "Drop mailing lists in favor of a forum software”:
http://markmail.org/message/dyfhqyuug7xgjru2
* Jeremie worked on a mail archiving app: see http://markmail.org/message/dyfhqyuug7xgjru2
and http://extensions.xwiki.org/xwiki/bin/view/Extension/MailArchive+Application
A - Evaluate and improve the Forum Application (
http://extensions.xwiki.org/xwiki/bin/view/Extension/ForumApplication) and
use our own Dog Food.
B - If it is too costly, use any PHP forum that the Open Source world have.
phpBB for example is the common choice. However, it does not centralize
everything but it replaces the couple ML/Markmail and these tools are very
well-known.
C - Use JIRA tickets with a certain label for development discussions
because sometime the debates are spread between issues and threads, so it
could be better to have everything directly on the issue. (FTR I don't
think JIRA is the right tool for that but I wanted to list all options).
This is not an action plan but a first step in that direction. Let me hear
what you think.
My global POV (will reply with more details later on, need to finish something first):
We need a solution that’s not too costly (in term of development, maintenance, etc). At
first sight, I’d say it’s not our objective to develop a forum (nor an issue tracker) and
whatever solution we develop would definitely not be a best of breed solution. Personally
I like best of breed solutions.
Note that In the other threads I mentioned there were other solutions proposed such as
stackoverflow, etc.
I think the first thing to do Guillaume is to read the previous threads and come up with a
design page on design.xwiki.org about the features (user stories, use cases) we’d like to
have in a our tool to replace the ML.
Thanks
-Vincent
Thanks,
Guillaume