TEI Technical Council Minutes 2015-10
TEI Technical Council Agenda and Minutes, 2015-10-28
Present:
-
Syd Bauman (SB)
-
Lou Burnard (LB)
-
Hugh Cayless (HC)
-
James Cummings (JC)
-
Stefanie Gehrke (SG)
-
Martin Holmes (MH)
-
Stefan Majewski (SM)
-
Paul Schaffner (PFS)
-
Peter Stadler (PWS)
-
Raffaele Viglianti (RV)
-
Fabio Ciotti (FC) - from 4 pm Sun
GUESTS ON WEDNESDAY MORNING:
-
Elisa Beshero-Bondar (EBB)
-
Antonio Rojas Castro (ARC)
-
Martina Scholger (MS)
Sunday 25 October, 09:00–17:00
venue: Mama Shelter, Atelier 3
SSID: Mama Shelter, login: Visitors; 5821191761
09:00–10:30: Agenda setting, etc.
10:30: Coffee Break
13:00–14:00: Lunch at hotel
Items for discussion
-
What is wrong with anyXML, and how can we fix it? (RelaxNG, Jing and @xml:ids)
-
Content model for macro.anyXML has not changed. Nevertheless, the introduction of <xenoData> has caused a problem to occur. The macro has been moved from tagdocs to tei module, which may be one reason. We need to find out whether the move to another module caused it, or its use in <xenoData>. It’s possible that the order of modules in an ODD may cause this; if so, we could constrain that with Schematron. HC has turned off one of the tests in the build process to avoid the problem for now. Action on LB to do testing and report back to Council. due 2015-11-25
-
Action on HC to figure out what specific test caused the problem and why. Also to create an Issue for this. Done.
-
For Schematron to check ID uniqueness, see wiki example
-
-
Release scheduling:
-
We don’t have to release on schedule, perhaps we should not feel forced to have a release when we’re still discussing issues? (but how does that pair with our desire to always have a release some time before members meeting)
-
“release-candidate”? or work on three levels in github : a/branches - b/development (all work there) - c/master
- git fetch + git merge https://git-scm.com/book/en/v2/Git-Branching-Branches-in-a-Nutshell
-
What is a release?
-
Guidelines and Schemas
-
Oxygen Packages
-
Debian Packages? Which ones?
-
HC suggests a new process taking better advantage of natural git methods of working. This would involve creating a new development branch, which would be where we branch from and merge to, and then build the development branch on Jenkins. The master branch (renamed to release?) would be where releases come from; merges are made from development to master before release. We should make the development branch the default on GitHub, and document this in the readme document.
-
Decisions after discussion:- copy current master branch to new dev branch
-
make the dev branch the default (on GitHub)
-
master branch is our new stable branch (aka release)
-
Ticket is closed when in dev branch
-
whole dev branch is merged into master once ONLY for release and tagged
-
Jenkins monitors dev branch.
-
When a tag is created for a release candidate (with a version number ending in e.g. rc1 for “release candidate 1”), the zip of the products built by Jenkins is pulled down and pushed to tei.github.io, where it is public and we can ask for feedback.
-
No one pushes to the Master unless making a release.
-
-
Adding and improving release processes and documentation
-
Jenkins: Victoria and Oxford or tei-c?
-
Or Amazon AWC? It’s a cost that the board would have to approve. Pricing: https://aws.amazon.com/ec2/pricing/ though note the costing is dependent on the number of back builds we keep, etc.
-
We need a stable URL for the Jenkins server, wherever it is, so that this can be changed once for all in all the spec files.
-
Consensus seems to be that we should get: jenkins.tei-c.org (redirected at tei-c.org to wherever it needs to be whether a local hosted copy or AWS or whatever)
-
Action: JC to get jenkins.tei-c.org (as soon as Martin configures his server to have it pointed at) also jenkins-backup.tei-c.org for another server.
-
Action: PWS will set up a Jenkins at upb.de; done.
-
SM: Travis-CI? any experiences. MH we are currently hitting the 30Min limit. SM could do a Tue hackathon on this. Would need reduction of build time because of limits.
-
-
Stylesheets management
-
What parts of the Stylesheets repo should Council be responsible for?
-
LB argues: All of it (all needed for co-products like guidelines)
-
prioritize : docx; pdf; html ; odf
-
develop workflow for stylesheets
-
-
Suggestion that “every issue should have an Issue” (e.g. both in Guidelines and Stylesheets if it affects both?)
-
Branch workflow in Stylesheets repository: we’ll use the same “dev” and “master” 2-main-branch system we decided upon for the Guidelines, above.
-
HC to send e-mail to non-Council Stylesheets contributors [turns out the only contributor with push access is Magdalena Turska, who's about to join Council anyway]
-
HC to actually change repo to the 2-main-branch system [done]
-
-
Debian packages?
-
several packages involved: stylesheets, Guidelines, etc.
-
Need a stable location for the packages, e.g. packages.tei-c.org (could then transparently to users be hosted at any of our institution).
-
Need a TEI gpg key for signing debian packages
-
list of packages, status: https://docs.google.com/spreadsheets/d/1w3ObgtqxpwYLrfaF-N0SPTt5X3ITmg2iLa5ZevAg3eE/edit?usp=sharing
-
Action on SM, SB, and MH to learn how to build & sign packages, and set up a repository, with a view to replacing the existing repository with a new one and creating a new build/sign/release process. Also to investigate if/which old packages could be dropped (TEI-P4 f.ex). Investigate whether we can use PPAs; By 2015-11-30. This should be tracked with an Issue.
-
-
Jenkins setup:
- Action on JC and MH to set up virtual hosts on their Jenkins machines for jenkins.tei-c.org and jenkins-backup.tei-c.org. Action on JC after this is done to ask Ian Rifkin to set up pointers on tei-c to point the first to the Oxford server and the second to the Victoria one. BY 2015-11-30.
-
Year-end release
-
Pure ODD
-
Processing Model Documentation
-
TEI Simple
-
Consensus: Release when ready but would be preferable to release Pure ODD and PM together.
-
-
Git / GitHub tutorial?
- Too many labels on GitHub. Action on RV to review, report to council, and clean up (or mentor new council member). Between election and Jan? Or do we feel it’s needed sooner? By 2015-12-25
-
Ticket Review
- BUGS
Monday 26 October, 09:00–17:00
venue: L'École normale supérieure de Lyon, Institut français de l’Éducation, meeting room 3 (conference venue page)
09:00–12:30 Ticket Review
12:30–13:30: Lunch (delivered)
13:30–17:00: Ticket Review
Tuesday 27 October
HACKING DAY: Some TEI Council members sat in a computer room at ENS and worked on their tickets and other TEI-related activity.
Wednesday 28 October, 09:00–12:30
venue: ENS again; guests (listed above) present
from 9:15
-
Round of introductions
-
HC introduces minutes document and general council workflow
-
Tickets
- Vote on @source in att.global.responsibility carried in favour against all sensible reason.
-
Stylesheets tickets
- Action on all: Please look at Stylesheets issues and choose a ticket or two by 2015-11-08; those tickets not assigned will be handed out at random (to MH, JC, SB, LB, RV, PS, HC, SM, keeping in mind those that have already taken tickets) by Chair on or about 2015-11-09.
-
TEI Simple
-
Action on all: Read docs pointed at below by 2015-11-12, and if you don’t want to actually edit it send comments to list or LB.
-
PM stuff: http://htmlpreview.github.io/?https://github.com/TEIC/TEI-Simple/blob/master/tei-pm.html (for the --summaryDoc version or for the full odd version see) http://htmlpreview.github.io/?https://github.com/TEIC/TEI-Simple/blob/master/tei-pm.odd.html
-
-
Pure ODD
-
Main job left is getting the display to look right, which is on HC.
-
LB asks that people check out the branch and test things
-
-
Next Release
-
Year end or wait until Simple-PM and Pure ODD are ready?
-
Test suite for processing model needs to be understood: we would like to ensure there is a test suite that properly exercises the structures and capabilities of the PM
-
JC believes that SPQR built these tests into the development of Simple. The https://github.com/TEIC/TEI-Simple/blob/master/build.xml has a ‘pm’ target (as does the Makefile) which runs a series of test files through the PM.
-
Action: JC to make a branch off of LB’s pureODD branch, add new tests there.
-
-
-
anyXML
-
Guidance on this problem belongs to the tagdocs chapter (specifically, a new para at the end of http://www.tei-c.org/release/doc/tei-p5-doc/en/html/TD.html#TDeg) , perhaps using the svg: example, linked to from a note in the header chapter (re: xenoData where the problem could conceivably arise, whereas egXML is a more likely locus for the problem.)
-
explanation should include a sample of Schematron that can be used for ID uniqueness checking
-
-
Meeting adjourned 11:31Z, with Council expressing its best wishes to Sebastian Rahtz and his family