Input for ODF plugfest/TC call

Tue Apr 24 22:58:02 BST 2012

Hi Jos,

thanks a lot for picking that list up as much as possible!

Am Freitag, 20. April 2012, 12:03:01 schrieb Jos van den Oever:
> On Thursday 19 April 2012 13:48:52 PM Friedrich W. H. Kossebau wrote:
> > so here my list of issues that I have had with ODF so far. Please consider
> > putting them on the table for the plugfest as well as for the TC call :)
> 
> I have brought up some points in the discussion.
> 
> > A) no official test suite
> > B) spec vs. reality: how to handle files from broken ODF producers
> > C) no feedback on office-comment list
> > D) handling of custom shapes just a hack ATM?
> > E) things I would have posted to the office-comment list if it felt live
> > 
> > 
> > A) No official test suite
> > 
> > As discussed on the ml, would be really good to have that, similar to what
> > there is for SVG at least.
> > The "OpenDocument Fellowship Test Suite" as used e.g. for
> > http://officeshots.org/galleries/view/opendocument-fellowship-test-suite
> > might be a start.
> 
> This is a common problem. KDE has a repository. There are some collections
> of files. A common repository would be a good idea. I've grilled the
> plugfest organizers on creating a repository. The reaction to this, which
> is the same for a long time, is to bring up a lot of potential problems
> instead of actually starting something. Currently the plugfest uses
> 'scenarios' which are not automated but require people to go through them.
> This is very inperfect and should be upgraded and I will continue to press
> for a plugfest-NG which has more automated scenarios.

I can only repeat myself and say that IMHO that is so needed, this makes you a 
hero :)

If you need volunteers to create some test files, please remember to ask me to 
help with that.

> > B) Spec vs. reality: how to handle files from broken ODF producers
> > 
> > E.g. LO (and so might OOo) do not meet the ODF 1.2 spec, section 19.228
> > draw:transform, where it says skewX/skewY should be given in degrees.
> > Instead they store the angle value as radian (and in clockwise direction)
> > (bug filed for LO as https://bugs.freedesktop.org/show_bug.cgi?id=48342).
> > And Calligra follows that behaviour.
> > 
> > Things like this screw the ODF spec. It needs to be handled.
> > But how?
> > * Fix ODF producers to produce spec-complying documents, and have
> > consumers
> > detect the producer version and handle the read values by the version?
> > * Adapt the spec to reflect reality?
> > 
> > Anyway, the current behaviour of the spread ODF producers IMHO should be
> > added as note to the spec, so implementors of ODF consumers know what to
> > expect.
> 
> I did not get a chance to bring this up.

I see that your test files you posted also to the office-comment ml are about 
skewX/skewY, so the problem how to deal with existing non-spec-complying docs 
should hopefully discussed as a follow-up.

> > C) No feedback on office-comment list
> > 
> > The official comment mailing list feels like writing to /dev/null. While
> > there is an automated monthly reminder which has this section
> > --- 8< ---
> > 3)  The TC has agreed to consider all submitted comments, regardless of
> > when they were submitted.  Although you may not receive an immediate
> > response from the TC, be assured that your comments have been archived on
> > the list, and you can track their resolution via JIRA, as explained below.
> > --- 8< ---
> > there has not been any official reply from the TC or any JIRA related
> > follow- up, at least visible in the the archives for the emails since Jan.
> > 1st 2011 and personally on the emails I posted there the last months.
> 
> We had a long discussion on this topic. Within ODF the workflow is such that
> publicly discussing the comments is discouraged. This means that there is a
> large delay between submitting a comment and it being handled by the TC
> members and even getting a reply back to the commenter.

Okay, good to see that posting to that list still seem useful. Just missing 
out any kind of non-automatic feedback in 1 1/2 years to show that actually TC 
humans are reading it really needs to be improved. As one usually first reads 
the archive to see how the common posting style is, and seeing no human TC 
reaction makes this list look a dead-end.

> > D) Handling of custom shapes just a hack ATM?
> > 
> > The current storage looks like a hack to me, it is no-where specified in
> > the spec, is it? And it does not survive a round-trip to a ODF
> > consumer/producer which does not support that custom shape. Which I
> > consider a fail :(
> 
> Custom object types is a recurring problem that comes up as a scenario every
> time and never works properly across all the products. How to save unknown
> objects back when saving is indeed not specified. The storage in a zip file
> is documented, that is not a hack.

Cmp. my reply to Thorsten here.

> > E) Things I would have posted to the office-comment list if it felt live
> > 
> > 1. line end markers are shitty
> > * no support for markers which are not centered at the Y-axis
> > 
> >   (think half-headed spear arrows)
> > 
> > * markers do not adapt to the stroke width -> ugly unless certain width
> > 
> >   while it's nice to be able to freely define the shape of a marker
> >   this does not work with the typical arrows.
> >   given that there is a set of commonly used arrow types it would be
> >   better
> > 
> > to hardcode identifiers for these types and have the ODF consumers care
> > for the actual rendering themselves, so being able to take care of nicely
> > integrating with the line stroke
> > 
> >   would also solve the problem with translations of the arrow/marker types
> >   the free definition of marker shapes should be still available of course
> >   or the stroke width should be available as parameter somehow to
> >   automatically the marker width, thus the marker fits
> > 
> > * only width stretching of markers possible, but not independently of
> > length * not possible to use multiple colors or fillings, i.e. complex
> > shapes e.g. the markers with "holes" in it are ideally filled with e.g.
> > white, not transparent. Or think of a pointing hand as arrow. Why not make
> > this possible?
> > 
> > 2. fillings do not support complex patterns, like SVG does
> > 
> > 3. spec misses to define for angle parameters the direction of the angle
> > 
> >    (clockwise or counter-clockwise)
> > 
> > 4. no parameter to say if fillings should be below or above outer line
> > 
> > In general it might also be a good idea to provide coders of import
> > filters
> > a way to collect all things from other formats that cannot be mapped to
> > ODF. I would have guessed office-comment list is for that, but well, if it
> > is it needs improvements to encourage people to commit there.
> 
> The comments mailing list is only for comments about the contents of the ODF
> standard. It is not meant to discuss interoperability with other standards.
> Thre is a separate TC for interopability:
>   http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=oic
> 
> In short: I a equally frustrated with the abundance of talk and a lack of
> actual work. The best thing to do is the common KDE approach: who codes
> decides. To move the TC forward, it is best to join and make things happen.
> 
> In regards to the test sets: I would really prefer to havea test set that
> easily calculates a score for each product.

Yes, would be good to have something that can be estimated automatically and 
mapped onto a single metric. But how would you calculate scores for the 
rendering? Can we expect pixel perfectness? And how to create the reference?

> That is however quite some work.
> We have some tests in calligra, that would be better to have outside
> Calligra.
> 
> A good idea might be to organize a sprint to make such a system. The sprint
> should have coders that are involved in the unit testing of the different
> products and the goal should be to make a repository of tests that can be
> automatically evaluated, for example by convering documents to pdf and
> analyzing the picture or round-tripping and evaluating the result. For
> calligra, I already made some tests like this written in a way that is
> calligra-independent for document import [1]. We could make similar tests
> for round

If there is some work to delegate, pass some over here.

Cheers
Friedrich