Yesterday I ran a triage with Tomasz Finc, PediaPress (PP), and members
of the community on the Collection extension -- the tool that WMF sites
use for creating PDFs as well as ZIM (http://openzim.org/
While Tomasz and I would have liked to get more MW developers to pick up
bugs, I was happy that to actually have some feedback from the
PP developers that will help us point to easier problems to
solve in the extension.
Also, due to personnel changes at PP, no one there had SVN commit
access. We were able to begin the process of resolving this and,
hopefully as a result, we'll get problems with the extension addressed
much more quickly.
You can find the lightly commented etherpad for the triage here:
-- PDF generation does not support
Complex Script Wikis
We started off talking about several bugs that can all be summed up
Since we support several different languages with
not-very-well supported PDF creation tools and some without PDF
creation support at all, we are pushing the limits of the current
PDF creation tool.
After we discussing this, Volker Haas (a PP developer) said that the
underlying tool, ReportLab (http://www.reportlab.com/
), was being
pushed to its limits and needed to be "rewritten from scratch".
PP doesn't have the resources to work on this right now, so I
updated the bug and assigned it to Tomasz so that he could begin to
figure out how we could solve it.
-- <noinclude> showing in PDF
After input from PP, this bug looks like a relatively easy
to fix on the MW site. Sumana tagged it as such.
-- Collection extension should not
add chapters in reverse order
-- The path to images
Chris Kepper (a PP developer) said he would look at these, but also
confirmed that #28118 should be easy.
-- Collection extension should
place time stamp of revision extracted into the offline file
Jessie Wild (Special Projects Manager in the Global Development at
WMF) requested this so that it would be easy to date the time a
collection was made.
The PP developers also confirmed that this one would be easy for a
MW developer to pick up -- and it doesn't involve PDF creation,
making it a particularly juicy target.
-- Creating a PDF with collection
extension does not render the <pages> tag hook from proofread page
At first, I didn't understand why this was needed. Luckily
helderwiki from ptwikibooks was in the triage and was able help me
understand its real need.
Ralf Schmitt (a PP developer) said he would work on this if time
-- Cascading page protection for
/Print template subpages
Since this one can be solved by any MW programmer (but it involves
hooks and permissions, so I hesitate to tag it "easy"), I unassigned
it from the PP developers.
Ralf Schmitt suggested that some of the concerns about hijacking a
page could be addressed by having a "See what this page looks like
-- Collection extension needs some
way to credit original authors of a work
Since helderwiki (mybugs on bugzilla) was in the triage, xe took the
chance to champion this bug. There is no established established
way of tracking non-wiki authors in MW, so there wasn't anything
that the Collection extension could do here right now. helderwiki
tracked down Bug #27629 (https://bugzilla.wikimedia.org/27629
"Summary of page editors/authors") and added it as a dependency for
-- openZim export should not drop
There was some confusion about this one. I gave it back to Tomasz
so he could get some clarification from ErikM for what was needed.
-- Collection PDF generation
doesn't handle <ruby> elements
Can't test since URL doesn't work. Tagged "testme" and asked if
someone could attach a testcase to the bug for future reference.
-- Collection extension doesn't sort
chapters when the user clicks on "Sort alphabetically"
Volker Haas said this one was really a UI bug -- something the PP
developers were currently working on.
-- Collection extension should
accept lists of chapters generated by templates (e.g. expanding
templates of a collection page before loading the collection)
Resolved WONTFIX by Ralf Schmitt because of the complexity involved.
-- Name of excluded-in-print
category cannot start with a namespace name
Resolved WONTFIX by Ralf Schmitt since everthing was working as
-- collection contents lost when
only loading js via https
Thanks to all who participated. There are a couple more followups on
this triage, but those will wait for another email.