Opencast Proposals
All important decisions for Opencast have to be made on list. To do that committers may send proposals (marked with
#proposal) to list on which other committers may then vote. Opencast uses lazy consensus meaning that no response
signals agreement. Apart from that committers may vote with:
+1yes, agree - also willing to help bring about the proposed action+0yes, agree - not willing or able to help bring about the proposed action-0no, disagree - but will not oppose the action going forward-1veto, disagree - opposes the action going forward and must propose an alternate action to address the issue or a justification for not addressing the issue
A proposal is accepted when no veto (-1) is voted within a time spawn of 72 hours (not counting weekends) after being sent to list.
Passed Proposals
Officially declare the Admin UI Facade as internal API for exclusive use by the module matterhorn-adminui-ng
Proposed by Sven Stauber sven.stauber@switch.ch, passed on December 16, 2016
Dear all,
I hereby propose to officially declare the Admin UI Facade as internal API for exclusive use by the module
matterhorn-adminui-ng.
Reason:
The Admin UI Facade is essentially the backend of the Admin UI.
While it would be technically possible to use this API for other purposes, this would introduce dependencies to
components other than the Admin UI.
Allowing such dependencies to come into existence would cause changes to the Admin UI Facade to potentially break
other (possibly unknown external) components.
Hence, we would need to announce, coordinate and discuss changes to this API to not break dependencies to components
we potentially don't even know.
This would unnecessarily slow down the future development of the Admin UI.
In addition, Opencast 2.3 introduces the External API which has been explicitly designed to meet the requirements of
an API used to integrate other components.
Changes needed:
The documentation needs to reflect that the Admin UI Facade is an internal API that will be changed without prior
announcement whenever needed without respecting dependencies other than the Admin UI itself and therefore people shall
not use this API for integration purposes.
Best,
Sven
Opencast Next: Code Cleanup
Proposed by Lars Kiesow lkiesow@uos.de, passed on Thu, 7 July 2016 15:21:19 UTC
Hi everyone,
a while ago we discussed on the technical meeting that we would like to
remove some old code from Opencast since these parts do not work
properly (sometimes not at all) or are unused.
Why cleaning up? To name some reasons:
- Less code to run (less memory, faster start-up)
- Less things to compile (faster build)
- Less dependencies
- People do not accidentally stumble upon broken things
- Less work for maintenance
And now here is what I #propose to remove and a reason why I think this
should be removed. I already took the comments people made in the first
draft [1] into account, although I still dared to include the two last
items but this time, hopefully with a convincing reason for why they
should be removed.
1. Old Administrative User Interface (matterhorn-admin-ui)
The reason for this should be obvious: We got a new one. The old one
has not been tested for the last three releases, is not linked
anywhere anymore and is partly buggy due to changes to Opencast. To
maintain two interfaces for one thing do not make sense.
2. Hold-state Workflow Operations
These do not work with the new interface any longer and the concept
has since been replaced by the actions you can perform on archived
material.
3. CleanSessionsFilter
Old temporary bug fix. For more details read the thread on our
developer list.
4. Republish Workflow Operation Handler
It can be removed since it has been replaced by a flag on
the publish operation in 2.x.
5. Old workflows + encodings
We got new ones. These were only left because of the old ui.
6. Old player (Flash in engage ui)
Flash is dead. We have the new player and Paella.
7. Most of shared_ressources
Almost everything in here belongs to old user interfaces.
8. matterhorn-engage-player
This is the old player Flex project. Iam not even sure it can still
be compiled.
9. matterhorn-test-harness
Old integration tests
10. matterhorn-mediapackage-ui
Old UI ressources
11. matterhorn-manager-*
Old, outdated configuration modification via web ui. This was never
used and would need a major update to get it working again at all.
12. matterhorn-load-test*
Some tests. I have never seen them executed by anyone.
13. matterhorn-holdstate-workflowoperation
Workflow operations requiring a hold state which does not exist
anymore with the new admin interface.
14. matterhorn-deprecated-workflowoperation
The name says everything. This includes the download DVD operation.
15. matterhorn-annotation-*
This should not work with either of the current players anymore.
16. docs/jmeter, docs/scripts/load_testing
Configuration for a performance testing tool. Not used for a long
time and not up-to-date.
17. Everything unused from:
https://data.lkiesow.de/opencast/apidocs/deprecated-list.html
E.g. FunctionException and ProcessExecutor(Exception)
18. matterhorn-webconsole
Karaf comes with a web console. We do not use our old implementation
anymore.
19. matterhorn-mediapackage-manipulator
Rest endpoint for media package manipulation. It's not used anymore
except by components to be removed.
20. matterhorn-search-service-feeds
Broken implementation for RSS/Atom feeds
21. matterhorn-caption-* and embed operation
Service for converting different subtitle formats and operation to
embed these subtitles into the media files. This is *not* player
caption support. If required, FFmpeg can be used for conversion
between several subtitle formats. Asked on list [2], no one uses
this.
As indicated before, points 20 and 21 had some comments for leaving them
in which did not convince me to not propose this. “Instead of removing
it, fix it” is an easy thing to say but sadly requires ressources.
Keeping it, announcing it as features and then tell people that it is
not working only afterwards is a bad thing and I would like to avoid
that.
Note that all the code is still in our history so that we loose nothing
if we want the old code back.
Please feel free to indicate if this action is fine for you or if you
want to keep some of the marked code. Please provide a reason if you do.
Best regards,
Lars
[1] http://bit.ly/28YOEZ1
[2] http://bit.ly/28Ztlt8
This proposal has passed with these additional corrections:
Hi,
we discussed this on today's technical meeting and I'm slightly
changing the proposal:
20. Let's remove matterhorn-search-service-feeds only after September
1st which is a realistic time to get things into the next Opencast
release. If someone has fixed the issue by them, we will, of
course, keep it.
This change takes into account that some people have said they are
interested into fixing that module, but will make sure that it's
removed if no one fixes it to not have an advertised but broken
feature.
21. I will be looking into adding subtitle support in a sensible way
before removing the matterhorn-caption-* modules or at least
clarify if they can still be used.
Regards,
Lars
Hi James,
a couple of days, I talked to someone saying that he will soon provide
a patch adding exactly this functionality. The holdstate operations are
definitely broken due to their UI.
My suggestion for a compromise here:
- Remove them if that patch for archiving the options is released
- Remove them if no one fixes them in time (September 1st) for 2.3
If you want to bring them back later, we always keep the code in our
history.
Regards,
Lars
> Hi,
> I would like to keep 2 and presumably 13. Both Manchester and AFAIK
> Cape Town have use cases for hold states since there is still no
> mechanism for passing WF configuration options from one WF to another.
> Regards
> James
The patch has already been published.
Opencast Community Repository Owners
Proposed by Lars Kiesow lkiesow@uos.de, passed on Fri, 13 May 2016 18:41:52 UTC
Hi,
today, in the technical meeting, we shortly discussed how to handle
requests, problems, etc regarding the other repositories we are hosting
under the umbrella of the Opencast community:
https://bitbucket.org/opencast-community/profile/repositories
While we have people who care about the official Opencast repository as
well as rules about what may be merged, who may merge things, … we do
not have that for other repositories and for some it's very unclear.
That is why I would like to propose that every repository under the
umbrella of the Opencast community needs to have a “project owner”
being responsible for that repository. Usually it should be the one
requesting that repository, but of course it can be someone else known
in the community.
I would also like to propose that if there is no one willing to take up
the responsibility to take care of a repository (ownership) if an old
owner leaves, the repository should either be removed or marked as
deprecated and moved to a separate section if so requested.
Finally, I would like to propose that we use the new “project” feature
of BitBucket to group the repositories into the groups:
- Opencast
- Contrib
- Adopters
- Deprecated (<- to be created if needed)
Currently, all repositories are in one big project.
Regards,
Lars
Rename Opencast Mailing Lists
Proposed by Lars Kiesow lkiesow@uos.de, passed on Thu, 14 Apr 2016 00:00:00 UTC
Hi everyone,
traditionally, we have the three mailing lists:
- matterhorn@opencast.org (development list)
- matterhorn-users@opencast.org (user list)
- community@opencast.org (more or less announcements)
Recently, though, we have seen especially the last two list being used
for user questions and problems. That is not surprising as we dropped
the name “Matterhorn” and new users do not know what that the list
matterhorn-users is meant for questions about Opencast.
That is why I would like to rename these lists to
- dev@opencast.org or development@opencast.org (I prefer the short
name but don't have very strong feelings about that)
- users@opencast.org
- announcements@opencast.org
Together with the already existing security-notices list, this gives
these lists a very clear meaning. It would also have the benefit that
users only interested in general announcements could subscribe to one
list only which would likely be a very low-traffic mailing list.
Additionally, this would make it sufficient to send announcements to
one list, instead of sending it to all three lists.
To prevent general questions on the announcements list, I suggest we
grant posting rights to board members, committers or other people who
have or had a role in our community only. I don't think we need to be
too strict here but should make sure that people understand what this
list is for.
Finally, for the sake of our current members, I would suggest that we
forward the mails to the old addresses for at least until the end of
the year, if that is possible.
Best regards,
Lars
Documentation Pull Request Merge Order
Proposed by Lars Kiesow lkiesow@uos.de, passed on Thu, 25 Feb 2016 20:52:00 UTC
Hi everyone,
as discussed in this weeks technical meeting, I hereby #propose to
allow out-of-order merges of documentation pull requests in the same way
we have this exception for bug-fixes.
to be precise, I #propose to change the development process docs for
reviewing and merging [1] in the following way:
[old]
- Pull requests for bug fixes (t/MH-XXXXX-...) may be reviewed and
merged out of order.
[new]
- Pull requests for bug fixes or documentation may be reviewed and
merged out of order.
Regards,
Lars
[1] https://docs.opencast.org/develop/developer/reviewing-and-merging/
Removing instances of print statements with a style rule #proposal
Proposed by Greg Logan gregorydlogan@gmail.com, passed on Wed, 12 Feb 2016 12:00:00 UTC
Hi folks,
I noticed in a recently review that there are still System.out.println
statements in use in our codebase. I was surprised, because thought we had
previously implemented a checkstyle rule which would have banned those
statements! I hereby #propose that we implement the changes outlined in
https://opencast.jira.com/browse/MH-11222, and remove these statements in
favour of logger statements. I also propose that we add this rule to the
checkstyle ruleset so that we don't have to deal with this again going
forward. Proposal closes EOD 2016-02-03.
G
How to release a new Opencast version…
Proposed by Lars Kiesow lkiesow@uos.de, passed on Fri, 14 Aug 2015 12:54:51 UTC
Hi everyone,
serving as co-release manager for two versions of Opencast, I noticed
that our current release process has some aspects of the release defined
in a way that is more hindering than helpful and I want to #propose a
slight change to these recommendations.
I hereby #propose:
1. Get rid of the `master` branch, make `develop` the main branch.
2. Do not use the --no-ff flags for merges
3. Do not create versions/tags in a release branch. Separate them.
Reasoning:
1. The short explanation whould be: When did you explicitely checked
out `master` last time? People rarely do that. If I want a specific
version, I use the tag, if not I want the release branch or
`develop`.
If you think about it, then the whole reason for `master` in GitFlow
is to always provide the last stable version to users who just check
out the repository and do nothing else. The problem with Opencast is,
that we support multiple versions at the same time. If in a couple
of weeks 1.6.2 is being released, it is the latest stable. Is it? If
I check out `master`, however, I will still get 2.0 as we cannot
merge 1.6.x afterwards. While you can grasp the reasons behind this,
it is a bit confusing for users and it is much easier to just tell
them to use the tag to check out a specific version.
That it, if they do not use the tarballs from BitBucket anyway.
2. First of all, most people seem to be using BitBucket for
auto-merging and it does not use --no-ff. So we are not really
consistent anyway. Being consistent and using --no-ff would mean to
forbid the usage of the BitBucket merge.
Second, have a look at the confusing mess that are the current
branches (I tried to find something in the visualization a while ago
but gave up). It would be much cleaner to try using fast-forward
merges. So instead of using non-fast-forward commits I would argue
that we should instead try to use as many fast-forward commits as
possible.
3. Once we decided to have the tags in our branches like this:
---- A ---- B (tagged) ----- C ---- D -->
A is the commit containing the version that is decided to be
released. B is the tagged version. It is exactly the same code as A
except for the pom.xml versions that are modified. Finally C then
reverts B as the modified version should not be part of the release
branch, .... After C, the code is basically A again except for the
history (which we later need to merge which can be problematic). D
would then be the next “real” commit, meaning the next fix.
Much easier to handle would be the following structure:
---- A ---- D -->
\
B (tagged)
You do not have to revert that commit, you do not need to merge the
easily conflicting pom.xml changes and in the end, you would anyway
check out the tag using git checkout <tag> if you want that
specific version
Branching structure:
To have a complete overview, this is what the new branching structure
would look like:
develop --*--*--*--*--*----*--------*--------*---->
\ / /
r/x.y.z *--*--*---*--*--*--*--*--*---->
\ \
* x.y.z-beta1 * x.y.z-rc1
Regards,
Lars
Moving away from the 3rd party scripts
Proposed by Greg Logan gregorydlogan@gmail.com, passed by Fri, 24 Jul 2015 16:45:40 UTC
Hi folks,
As it stands right now we depend on the 3rd party tool script to
install a great many of our 3rd party dependencies. These are
utilities like tesseract, ffmpeg, sox, etc. This script is maintained
by Matjaz, in his own time. I'd like to take a moment to thank him
for a doing a great job on a particularly annoying aspect of
supporting our work! I know it hasn't been easy, especially
supporting vast number of different OS versions!
With the release of 2.0 I noticed that our 3rd party tool script is
becoming both a little out of date, and difficult to maintain. I took
a quick look around and it seems like *most* of our dependencies are
available from normal distribution repositories for Debian based
systems, and I'm told that there is a similar situation for Redhat
based systems. I am unsure of how many of our users are running
Matterhorn on Mac, but I would hope that our developers who are
working on Mac would be able to provide instructions and/or binaries
for those users. The only dependency where there might be a universal
sticking point is ffmpeg (due to patent concerns), however ffmpeg
builds a full static binary with each release, so I assume we can
either depend on this and/or cache them somewhere.
What this means is that we can potentially remove the 3rd party script
from our repository. I hereby #propose we find a way to do that,
which would remove the 3rd party script from the repository and
replace it with a number of new steps in the install documentation.
G
Status of youtube in 2.0 and #proposal to change the default workflow
Proposed by Rüdiger Rolf rrolf@Uni-Osnabrueck.DE, passed on Sat, 13 Jun 2015 14:15:55 UTC
Hi list!
There was some discussion in the DevOps meeting yesterday if the
Youtube distribution would work or not. I offered to check this.
The good news first: IT WORKS!
Just follow this manual and your Matterhorn - ups Opencast - is ready
to distribute to Youtube.
http://docs.opencast.org/r/2.0.x/admin/modules/youtubepublication/
The bad news: The default workflow definition does not really support
the publishing on Youtube, as only one video file could be published
by the current WOH.
https://opencast.jira.com/browse/MH-10920
The reason is simple and the fix would be too. But there are some
options to fix this:
1. Remove the option to distribute to Youtube from the default workflow
definition, as the complicated configuration would have to come
first anyway.
2. Only let "presenter" or "presentation" be published to Youtube. We
would need a new youtube tag and add this to the compose operation
and the youtube operation.
3. Introduce the composite operation to the workflow definition and
publish only the resulting single stream to Youtube.
4. Upgrade the WOH to support publishing of multiple files.
I would say that option 4 could be 2.1 goal, but not for 2.0.
I would #propose to go for option 1, as nobody can use Youtube
out-of-the-box anyway. And the admin could then setup an appropriate
Youtube workflow for their needs too.
Regards
Rüdiger
Episode DublinCore Catalog
Proposed by Karen Dolan kdolan@dce.harvard.edu, Passed on Sat, 30 May 2015 12:39:05 UTC
Dear Opencast-ees,
The following proposal addresses MH-10821[1]. An issue that exposes a
know long time ambiguity regarding metadata and the ingest service.
The reason that its a proposal is that it normalizes the handling of
inbound episode catalog metadata in the ingest service.
1) A new configuration parameter, boolean, for the Ingest Service. The
config param identifies if episode metadata precedence is for
Ingestee (i.e. Opencast system) or the Ingester (i.e. Capture
Agent).
For example: at our site, the scheduling entity is the metadata
authority. All updates are made to the Scheduling endpoint. The
Capture Agent always has stale episode catalog metadata. At other
sites, updates are made on the Capture Agent directly. The
community default can be for priority to the Capture Agent.
2) All Ingest endpoints perform the same consistent process to ensure
that an episode catalog will exist, manually or automatically
provided.
3) The process performs the following...
3.1. Gather data
- Check if inbound media package contain a reference to an Episode
DublinCore catalog and if that catalog contains a title.
- Check if the inbound media package contains a title attribute.
- Check if the Workflow service has a reference to the mediapackage's
Episode Dublin Core catalog
- Check if the Scheduler service retained a reference to the event's
Episode Dublin Core catalog
3.2. Use config param to prioritize action on acquiring an Episode dc
catalog for the media package
If Capture Agent metadata takes precedence:
- Take the inbound Episode dc catalog, if it exists
- Take the Episode dc catalog from the workflow service, if it
exists
- Take the Episode dc catalog from the scheduler service, if it
exists
- Create an Episode dc catalog from the title in the media
package,, if it exists
- Create an Episode dc catalog using a default title (i.e.
"Recording-1234556XYZ")
If Opencast metadata takes precedence:
- Take the Episode dc catalog from the workflow service, if it
exists
- Take the Episode dc catalog from the scheduler service, if it
exists
- Take the inbound Episode dc catalog if it exists
- Create an Episode dc catalog from the title in the media
package, if it exists
- Create an Episode dc catalog using a default title (i.e.
"Recording-1234556XYZ")
I'll start a pull for the above, and appreciate any thoughts.
Regards,
Karen
[1] https://opencast.jira.com/browse/MH-10821
Dropping Taglines
Proposed by Greg Logan gregorydlogan@gmail.com, Passed on Fri, 29 May 2015 16:19:09 UTC
Hi folks,
I hereby propose that we drop the practice of having taglines. I
propose this because we don't have a place in the new admin UI to put
them, nor have I ever heard any of the adopters make use of it. I know
we don't use it as a committing group, which means that *no one* is
using them.
G
Wiki Cleanup
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Fri, 24 May 2015 11:36:49 UTC
Hi everyone,
since we partly switched to our new documentation [1] I would like to
make sure that the old and mostly outdated documentation goes away so
that no one stumbles upon that. When I had a look at the wikis we
currently have I noticed that most of our 17(!) wikis have not been
touched in years and can probably go away.
Here is a list of our wikis and what I #propose to do with/to them:
Keep (maybe clean-up a bit):
- Matterhorn Adopter Guides
- Matterhorn Developer Wiki
- Opencast Matterhorn D/A/CH
- Opencast Matterhorn Español
- LectureSight
Export as PDF to archive the contents and then delete:
- Matterhorn Release Docs - 1.0
- Matterhorn Release Docs - 1.1
- Matterhorn Release Docs - 1.2
- Matterhorn Release Docs - 1.3
- Matterhorn Release Docs - 1.4
- Matterhorn Release Docs - 1.5
- Matterhorn Release Docs - TRUNK
Keep until 2.1 is out then export as PDF and delete:
- Matterhorn Release Docs - 1.6
Just delete:
- Analytic video annotation
- Infra
- Matterhorn Documents
- Opencast Community
Please let me know if you agree or disagree with this proposal.
Regards,
Lars
[1] http://documentation.opencast.org
Jira Clean-Up
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Fri, 8 May 2015 11:52:16 UTC
Hi everyone,
as discussed in the technical meeting, I hereby #propose:
The “Blocker” and “Release Blocker” severity status are more or less
redundant. As part of cleaning up Jira, let us remove the “Release
Blocker” severity in favor of “Blocker”.
As footnote, some statistics: Since the beginning of 2014, 70 Release
Blockers have been files in Jira while mere *8* Blockers have been
files.
Regards,
Lars
Opencast Documentation
Proposed by Rüdiger Rolf rrolf@uni-osnabrueck.de, Passed on Sat, 02 May 2015 14:43:28 UTC
Hi all,
Tobias, Basil, Lars and I discussed status of the current migration of
the Opencast (Matterhorn) documentation to GIT. We still see some open
issues that need clarification so we would like to propose the
following points:
*1. Formating and Hosting of the Documentation *
We want to use https://readthedocs.org to or a similar service create
a more appealing HTML version from the Markdown of the documentation.
The documentation will be versioned there so that for older versions
the documentation is still available. By default the "latest" version
is shown. The versions of the documenation will be generated based on
the release branches.
*2. Structure of the Documentation*
We see the documentation in*Git *separating into 3 sections:
- /Administration Guide/: with information about the installation,
configuration, customization and integration. This will be the part
of information by an administrator to setup
Opencast.
- /Developer Guide/: All information related to implementation
details of Opencast, so that this will be updated in a pull request
(API changes, module descriptions, architecture). The development
process documents should also go here as only committers usually
should change these.
- /User Guide/: Documentation of the (new) Admin UI that was already
started by Entwine and the Engage UI (especially Theodul Player).
This guide should only describe options available on the UIs.
Within the *Wiki* we still see the need for 2 sections:
- /Developer Wiki/: Proposals, working documents and meeting notes
will be kept here so that anybody can edit these. So information
not to close to any existing implementation that might still be in
a process of discussion can be found here.
- /Adopters Wiki/: This can be the place where adopters share their
best practises, configurations, hardware recommendations,
third-party software documentation etc. Again anyone can contribute
to this wiki.
The difference between the Wiki and Git is in the first line that the
Git documentation should become a quality assured ressource for
Opencast users. The Git documentation should be reviewed within the
release process and it will be part of the review process of a pull
request, to make sure that the needed documentation changes have been
contributed too.
The Wikis on the other hand should be a more open platform where
everybody can contribute and users might find cookbooks to enhance
their system, or they can share ideas.
So now we would like to get your opinion on this proposal.
Thank you,
Rüdiger
Requirement Specification
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Thu, 16 Apr 2015 15:55:31 UTC
On list or IRC we often see that people do not really know the current
requirements for a specific version of Opencast Matterhorn. Of course
there are the pom.xml files specifying internal dependencies, but there
is nothing for 3rd-party-tools, ...
It would be nice to add a file specifying these requirements in a
format that is easy to parse and can hence be used for automatic
scripts to generate dependency lists, ...
That is why I hereby #propose to add a requirements.xml file that
specifies the requirements for Opencast Matterhorn:
- Required tools including versions
- Which modules require which tools
- Which modules conflict with each other (negative requirement)
This is mainly what is not specified by the pom.xml files yet.
Jira Clean-Up (Tags VS Labels)
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Thu, 19. Mar 2015 15:43:20 UTC
…then hereby I officially #propose removing the labels from Jira.
For more details, have a look at the mail thread at:
https://groups.google.com/a/opencast.org/forum/#!topic/matterhorn/vIdWQkZmbdQ
FFmpeg Update
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Sat, 14 Mar 2015 22:12:18 UTC
Looking at the FFmpeg project for the last two years, you will notice
that they developed a pretty stable release cycle with a release of a
new stable version approximately every three month.
To stop us from having to propose an update again and again, I hereby
propose the following general rule for our support of FFmpeg:
A Matterhorn release will oficially support the latest stable
version of FFmpeg released at the time the release branch is cut and
all other FFmpeg versions with the same major version number released
afterwards.
For example, for Matterhorn 2 this would mean that we will officially
support FFmpeg 2.5.4 and all later 2.x versions like 2.6 which has
been released on the 7th of March or a possible 2.7 onece it is
released. We would, however, not necessarily support an FFmpeg 3 as it
*might* come with an interface change that *could* break compatibility.
That obviously does not mean that older versions of FFmpeg just stop
working. In fact, most parts of the default Matterhorn configuration
should at the moment still work with FFmpeg 1.x but we will not test or
fix compatibility problems.
Proposal Log
Proposed by Lars Kiesow lkiesow@uos.de, Passed on Sat, 14 Mar 2015 16:35:08 UTC
It would be wonderful if we had a central place to look up the proposals
that have passed.
That is why I hereby propose that:
- We create a proposal log in our new documentation containing all
proposals that have passed on list.
- A proposal will become effective only after it is written down in
that log. That should usually be done by the person who sent out
that proposal.
This will, of course, not affect the existing decision making rules
(proposal on list, marked with #proposal, lazy consensus after three
days, no -1, ...)