Since my last proposal didn’t get a consensus, let’s restart the discussion in more
general terms:
Current Strategy
=============
We defined it in the "xwiki core” thread (source:
):
"
Executive summary:
* Reduce the scope of all the code located in the xwiki github organization by only
keeping “core” modules
* A “core" module is defined by being a generic transversal module (i.e. that can be
used in lots of XWiki flavors, if not all). This is opposed to “vertical”
modules which are modules specific of a usage of XWiki.
** Examples of “core" modules: logging module, configuration module, distribution
wizard, annotations, active installs, one base flavor (the “XWiki
” flavor), etc
** Example of “vertical” modules: meeting manager application, blog application, FAQ
application, flavors (except the base flavor), etc
"
Right now the goal of the XWiki Platform is to provide the **most usable generic runtime
possible**.
According to this definition the Tour and CKEditor are both clearly platform modules.
Note that the goal of flavors right now is to provide verticals. It’s not to provide
additional extensions that make the generic XWiki runtime more usable.
Future Options
============
There are only 2 options that are different from the current one and that I can imagine:
A) Keep the Tour application and CKEditor extension outside of XWiki Platform. The
consequence is that the current strategy doesn’t apply anymore and we would need a new
definition of XWiki Platform. One that I could see would be “a minimally usable generic
runtime”, which is very undefined. For example, is the Annotation feature part of a
minimal distribution. Answer: no. Actually we could even question if the EM is needed for
a minimally usable runtime (first versions of XWiki were usable and didn’t have the EM).
Activity Stream? AppWithinMinutes? Chart feature? So we would need to define clearly what
we call a minimally usable runtime.
B) Move the Tour application and CKEditor extension inside XWiki Platform but start having
different version/release lifecycle for some extensions. If we do this for the Tour or CK
then there’s no reason not to do it for other extensions and very quickly we go back to
what we were doing 6-8 years ago where it was a major PITA to do any release and to
validate what extension version was working with what other extension.
For me B) is not a good option and neither if A) unless we find a way to clearly define
it.
My worry for A) is that it’s easy to say that we want a minimal runtime and this would
mean moving a LOT of modules outside of platform and into contrib, which would mean a lot
more work to maintain/release them (each module would have its release cycle). Releasing
platform would becomes simpler (less stuff to build) but release the default flavor (in
contrib) would become a major pain. And also, who would do it since there’s no notion of
Release Manager for contrib. Any dev of contrib could release a new version of the default
flavor and thus change what the default XWiki runtime is. I feel we need to control what
the default XWiki runtime is and that should be the XWiki Core Dev Team who does this.
So at this point I believe that we need to keep the default flavor in platform and I
don’t see any viable option other than the current strategy.
@Denis: if you’re in favor of A) please let us know how you’d implement it in detail and
how you’d answer the challenges I raised.
Thanks
-Vincent
On 09 Jun 2016, at 11:01, Vincent Massol
<vincent(a)massol.net> wrote:
Hi everyone,
Thanks for the replies. I’m listening of course to everyone and I’ve tried in this mail
to take all answers into account.
First, let me state our current strategy and an alternative that I’ve been thinking about
this morning under my shower ;)
Current Voted Strategy
==================
* Deliver an XWiki Runtime that is the best possible generic runtime (i.e. most usable,
most useful).
* As a consequence, remove all modules that vertical modules (i.e. that are clearly not
useful to all flavors), such as FAQ, Blog, etc. Move them to xwiki-contrib
* I want to stress out that the current voted strategy is not to produce a minimalist
runtime
New Strategy Proposal
==================
I’ve tried to reconcile all the use cases listed in this thread before and I hope this
proposal could be a good middle ground. In any case I found it worth debating to see if it
could work.
Also note that one aspect that we must not forget (and that led to the last proposal I
sent on this thread) and that people tend to forget, is the time it takes to support
various versions of XE in an extension and the manpower that exists in the xwiki community
(don’t forget that everything we do is a tradeoff; if you support another version of XE in
an extension, it means you’re not coding an important improvement or fixing an important
bug in the platform).
So here’s the idea:
* Change the purpose of the XWiki Github organization from the voted one described above
to be: Provide a minimalist runtime.
* Since working in this direction will not happen overnight, the idea would be to very
slowly take out modules, starting with obvious ones.
* The issue that this strategy raises is that users will not get a good user experience
since lots of things will be lacking and this is where my new idea fills the gap:
** The first time (or whenever you upgrade) your run the XWiki Runtime (be it whether
your run the HSQLDB/Jetty packaging or any other packaging) you get a Configuration
wizard
** This Configuration Wizard suggest some recommended extensions that the XWiki Core Dev
Team hand-pick. We would start with 2:
*** Propose to the user to run a Tour to learn how to use XWiki (it would install the
Home Page Tour which depends on the Tour app)
*** Propose to the user to install the CKEDitor WYSIWYG editor (by default we would only
propose the wiki editor - We’ll need to get rid of the GWT editor, probably make it an
extension)
Pros:
* The XWiki Core Dev Team continues to work on core stuff and as time progresses we move
out non core stuff
* This allows more people to contribute to the non-core stuff in the community
* We control which extensions we want to recommend and thus we can always only take the
very good ones and thus control the quality of the initial user experience
* We get a mechanism allowing our users to get non-xwiki core dev team-supported
extensions into the runtime (thus providing a good user experience) while not bundling
them into the default XWiki runtime flavor.
Cons:
* The Tour and CKeditor extensions would still incur a higher cost of support/maintenance
(but since they’ve already done the code, it’s marginal for the future and they’ll be able
to abandon support for XWiki 6.x soon IMO too - Basically they probably only need to
support 2 or 2.5 cycle versions).
<similar idea>
Ludo mentioned (and I agree with him) that it would be nice to be able to provide Demo
content in the wiki so that users who want to test drive XWiki can do so with existing
content and more clearly see and understand the advantages that XWiki brings. For this,
I’d propose to create a Demo Content extension (some AWM app + some blog posts + etc) and
once we have it, recommend it on the Configuration Wizard.
</similar idea>
WDYT?
Thanks
-Vincent
> On 08 Jun 2016, at 17:57, Denis Gervalle <dgl(a)softec.lu> wrote:
>
> Well, very sorry to drop in so late in this discussion, but it was not
> obvious from the thread subject that your were discussing a major subject.
>
> IMO, moving application that works currently on 6.x to the core, has no
> benefit for our users, it just introduce restrictions. It does not have any
> benefit for us either, it just require more backports. I do not understand
> this move at all for application that are not minimal requirements. I do
> not understand your point Vincent when you say that these applications are
> horizontal and obviously part of platform according to your "Executive
> summary".
>
> Regarding the tour application, it is not require at all, it is just a nice
> helping tool that we want to ease newcomers, but experienced user will
> never need it. It could be exchanged for an alternative, and it is exactly
> the same kind of application than the blog that we are moving out.
> Regarding the CKEditor, do we consider that a WYSIWYG editor is required
> for a wiki to be a wiki ? IMO, WYSIWYG editor is not a requirement to use
> the platform, it is nice to have, but not required. I have use it very
> sparsely until now, and not having it would not have change much for me.
>
> So, I currently do not see any benefit of moving these modules to platform,
> since these are already well living in contrib.
>
> Your other point about reducing platform to the minimal runtime would cause
> platform to reduce to EM does not really looks like something that will
> happen. In theory, you are right, so XWiki would be even less featured then
> maven. But, I doubt you could reasonably use such a tools for anything
> useful. I doubt XWiki compare to maven. I doubt that horizontal module like
> security, logging, model, storage, etc… will ever be considered optional.
> Even a plain text editor is a minimal requirement to starts, else this is
> no more a wiki, and I even wonder what it is ? a tool that brings together
> arbitrary java module… looks weird. So no, the minimal runtime is
> definitely not just EM.
>
> So, I really wonder what is the direction we are taking. I will not stop
> you with a veto, but I have the strong feeling these decisions are wrong.
> For the principle of not depending on contrib for our default user flavor,
> exchanging the blog app with the tour app, this does not make sens for me,
> sorry.
>
> Thanks for reading.
>
>
>
> On Wed, Jun 8, 2016 at 3:45 PM, Vincent Massol <vincent(a)massol.net> wrote:
>
>> FTR I’ve discussed internally with Thomas, Marius and Anca and we all
>> agreed that it makes sense to move The Tour app + CKEditor to the platform.
>>
>> There are various reasons but a very important one is simply the manpower
>> that it requires to maintain extensions on lots of XWiki versions and
>> currently the active devs on xwiki are not enough to do that. This is the
>> reason we dropped this strategy in the past and decided to release the
>> whole platform together with the same version.
>>
>> As part of this the technical debt is being increased since supporting
>> several versions and old versions means doing hacks.
>>
>> If you see another possibility that doesn’t require more work please raise
>> it here.
>>
>> We need to progress and have CKEditor and Tour bundled in 8.2M2(which is
>> already started) and thus, barring any negative comments, we’ll start the
>> move next week.
>>
>> Thanks
>> -Vincent
>>
>>> On 07 Jun 2016, at 15:39, Guillaume Delhumeau <
>> guillaume.delhumeau(a)xwiki.com> wrote:
>>>
>>> It also means to move the tour application in that old branches too.
>>>
>>> 2016-06-07 13:59 GMT+02:00 Vincent Massol <vincent(a)massol.net>et>:
>>>
>>>>
>>>>> On 07 Jun 2016, at 10:27, Vincent Massol <vincent(a)massol.net>
wrote:
>>>>>
>>>>>
>>>>>> On 07 Jun 2016, at 09:37, Guillaume Delhumeau <
>>>> guillaume.delhumeau(a)xwiki.com> wrote:
>>>>>>
>>>>>> Moving Tour Application into platform makes sense to me (it
becomes a
>>>>>> critical component and deserves a proper support).
>>>>>
>>>>> For me, it’s really about the definition of what the XWiki github
org
>>>> represents. Right now with the new strategy == “Everything needed for
>> the
>>>> default XWiki runtime, a.k.a base/default flavor” (what we’ve been
>> calling
>>>> XE so far but that we’ll slim down a bit, for example by removing the
>> Blog
>>>> app and move it to contrib).
>>>>>
>>>>> Now we could still decide to have some flavor in contrib and have
the
>>>> tour app included in that flavor but not in “the default XWiki
>> runtime”. In
>>>> practice this would mean promoting this flavor instead of the
>> base/default
>>>> flavor. The question will arise anyway when we next talk about other
>>>> flavors that we may want to have in contrib such a KB flavor, workgroup
>>>> flavor, web flavor, etc.
>>>>>
>>>>>> However, the current
>>>>>> application supports XWiki >= 6.4.1. By moving it to platform,
we will
>>>> only
>>>>>> support the last XWiki version.
>>>>>
>>>>> This is a tough topic indeed.
>>>>
>>>> Actually in practice we would support not only the last XWiki version
>> but
>>>> also the LTS (i.e. 7.4.x + 8.x). If we wanted to support 6.4.x we could
>> (we
>>>> still have a stable-6.4.x branch ATM that we were supposed to remove)
>> but
>>>> it would mean changing our support strategy to support more branches…
>> and
>>>> it means supporting the whole platform for 6.4.x, not just one
>> extension…
>>>>
>>>> Thanks
>>>> -Vincent
>>>>
>>>>
>>>>> For the tour there’s the solution of keeping it in contrib and
>>>> introducing a flavor but for CKEditor it’s harder to justify that it’s
>> not
>>>> part of the base flavor IMO but maybe it’s possible and we would offer
>> only
>>>> the wiki editor in the base flavor. Of course we could modify our
>>>> functional tests fwk to support running on various versions of the
>>>> dependencies and have CI builds to ensure that an extension works with
>> all
>>>> versions but it’s not perfect and it would mean that for the first time
>> we
>>>> would have code in the xwiki github org that would not use the latest
>>>> APIs/latest JDK features.
>>>>>
>>>>> The other option is Marius’s, i.e. accept that we hand-pick some
>>>> extensions from contrib that we bundle in the base/default flavor such
>> as
>>>> the Tour app, CKEditor integration, etc. In this case, we would just
>> need
>>>> to redefine what “xwiki github org” means. Saying “core component” would
>>>> not be enough, it would needs a more precise definition.
>>>>>
>>>>> Interesting topic ;)
>>>>>
>>>>> Any other option that we have?
>>>>>
>>>>> Thanks
>>>>> -Vincent
>>>>>
>>>>>> 2016-06-06 15:31 GMT+02:00 Vincent Massol
<vincent(a)massol.net>et>:
>>>>>>
>>>>>>>
>>>>>>>> On 06 Jun 2016, at 15:24, Marius Dumitru Florea <
>>>>>>> mariusdumitru.florea(a)xwiki.com> wrote:
>>>>>>>>
>>>>>>>> On Mon, Jun 6, 2016 at 3:58 PM, Vincent Massol
<vincent(a)massol.net>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>>>
>>>>>>>>>> On 06 Jun 2016, at 14:50, Marius Dumitru Florea
<
>>>>>>>>> mariusdumitru.florea(a)xwiki.com> wrote:
>>>>>>>>>>
>>>>>>>>>> On Mon, Jun 6, 2016 at 3:09 PM, Alexandru Cotiuga
<
>>>>>>>>>> alexandru.cotiuga(a)xwiki.com> wrote:
>>>>>>>>>>
>>>>>>>>>>> Hello all,
>>>>>>>>>>>
>>>>>>>>>>> As it was decided already, a Homepage Tour
have to be
>> implemented.
>>>>>>>>> However,
>>>>>>>>>>> no option regarding the place where the Tour
Application should
>> be
>>>>>>>>> added as
>>>>>>>>>>> dependency was discussed.
>>>>>>>>>>>
>>>>>>>>>>> There are some possible options:
>>>>>>>>>>> 1) XWiki Enterprise
>>>>>>>>>>> 2) XWiki Platform Distribution
>>>>>>>>>>> 3) XWiki Platform Helper
>>>>>>>>>>>
>>>>>>>>>>> 4) Is there any option to have the Tour
Application as a part of
>>>> the
>>>>>>>>> Core ?
>>>>>>>>>>>
>>>>>>>>>>> What would be the best way to include the
Contrib applications in
>>>>>>> XWiki?
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On this topic (sorry if I hijack your thread) I
was wondering why
>>>> don't
>>>>>>>>> we
>>>>>>>>>> have dependencies from platform/enterprise to
contrib. We have
>> lots
>>>> of
>>>>>>>>>> third party dependencies, contrib could be
considered as such.
>>>>>>> Moreover,
>>>>>>>>>> we're in the process of moving non-core
(vertical) extensions out
>> of
>>>>>>>>>> platform to contrib. It would be a pity to move
something from
>>>> contrib
>>>>>>> to
>>>>>>>>>> platform and then back to contrib. I have the
same issue with the
>>>>>>>>> CKEditor
>>>>>>>>>> Integration extension. We want CKEditor as the
default editor,
>>>> bundled
>>>>>>>>> with
>>>>>>>>>> the default distribution, but do we need to move
it to platform?
>>>> Same
>>>>>>> for
>>>>>>>>>> the Welcome Tour.
>>>>>>>>>
>>>>>>>>> I’d personally not like this for the following
reasons:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>> 1) I like that the XWiki runtime is all released at
once with all
>>>>>>>>> extensions making it using the same versions and
verified to work
>>>>>>> together.
>>>>>>>>>
>>>>>>>>
>>>>>>>> XWiki runtime has lots of third party dependencies.
Bootstrap, Solr,
>>>>>>>> jQuery, just to name a few. I don't see how having
the source code
>> in
>>>> our
>>>>>>>> repo (platform) makes a difference at runtime when the
>>>>>>>> integration/functional tests verify they work together.
>>>>>>>
>>>>>>> Because they don’t! :) Just check any extension in contrib
and you’ll
>>>> see
>>>>>>> their func test (when they have some!) don’t test that they
work with
>>>> the
>>>>>>> latest version of XWiki…
>>>>>>>
>>>>>>>> 2) Support. The XWiki runtime is supported by the XWiki
Core Dev
>> Team.
>>>>>>>>> Extensions in contrib are not supported by the XWiki
Core Dev Team.
>>>>>>>>
>>>>>>>>
>>>>>>>> So the FAQ application you moved out of platform is no
longer
>>>> supported
>>>>>>> by
>>>>>>>> the XWiki Core Dev Team?
>>>>>>>
>>>>>>> Correct.
>>>>>>>
>>>>>>>> The extension page
>>>>>>>>
>>
http://extensions.xwiki.org/xwiki/bin/view/Extension/FAQ+Application
>>>>>>>> doesn't reflect this.
>>>>>>>
>>>>>>> I added my name to the list as a supporter. I’ve kept “XWiki
Dev
>> Team”
>>>>>>> because it's a past authors and it wouldn’t make sense to
remove it.
>>>> But
>>>>>>> yes it’s no longer officially supported by the XWiki Core Dev
Team.
>>>>>>>
>>>>>>> Note that e.x.o doesn’t say who maintains a given extension,
it just
>>>> says
>>>>>>> who participated to developing it ;) We’re currently missing
the info
>>>> on
>>>>>>> whether the extension is actively supported and by whom. FTR
>> Confluence
>>>>>>> does this with a “supported” label that you can hover over
and
>> provides
>>>>>>> info. For example:
>>>>>>>
>>>>
>>
https://marketplace.atlassian.com/plugins/nl.avisi.confluence.plugins.numbe…
>>>>>>>
>>>>>>> Thanks
>>>>>>> -Vincent
>>>>>>>
>>>>>>>> In addition xwiki-contrib is very open and anyone can
make
>>>> modifications
>>>>>>>>> there and quality is thus harder to guarantee.
>>>>>>>>>
>>>>>>>>> We defined the xwiki github organization as
containing horizontal
>>>>>>> modules,
>>>>>>>>> ie modules that can be required for any flavor and
both CKEditor
>> and
>>>> the
>>>>>>>>> Tour Application fit the need. By opposition to
vertical modules
>>>> which
>>>>>>> make
>>>>>>>>> sense only for some use cases (like the Meeting
Manager app) and
>> not
>>>> by
>>>>>>>>> default in XE. We have the option of having flavors
in contrib for
>>>>>>> those if
>>>>>>>>> we want though. For CKEditor it’s not a good thing
since we’d like
>>>> it by
>>>>>>>>> default.
>>>>>>>>>
>>>>>>>>> One alternative (which I’m not fond of at all) would
be to have
>>>> ckeditor
>>>>>>>>> as a separate git repo in the xwiki github
organization.
>>>>>>>>>
>>>>>>>>> Thanks
>>>>>>>>> -Vincent
>>>>>>>>>
>>>>>>>>>> Thanks,
>>>>>>>>>> Marius
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Thanks,
>>>>>>>>>>> Alex