Re: ISSUE-162 (profile discrepancy): profile discrepancy [DFXP 1.0]

On Fri, Mar 2, 2012 at 10:28 AM, Timed Text Working Group Issue Tracker <
sysbot+tracker@w3.org> wrote:

> ISSUE-162 (profile discrepancy): profile discrepancy [DFXP 1.0]
>
> http://www.w3.org/AudioVideo/TT/tracker/issues/162
>
> Raised by: Sean Hayes
> On product: DFXP 1.0
>
> From: Shaun McCance [mailto:shaunm@gnome.org]
> Sent: 02 March 2012 13:48
> To: public-tt@w3.org
> Subject: Discrepancy in profiles?
>
> Hi all,
>
> I've been working on embedding TTML in Mallard documents to provide
> subtitles for videos. See, for example, here:
>
> http://git.gnome.org/browse/gnome-games/tree/gnotravex/help/C/usage.page
>
> While working on profiles, I noticed these two specifications of the
> dfxp-presentation profile are different:
>
> http://www.w3.org/TR/ttaf1-dfxp/#profile-dfxp-presentation
> http://www.w3.org/ns/ttml/profile/dfxp-presentation
>
> The standalone file requires 15 features not required by the profile
> in the recommendation. The standalone file also uses a base attribute,
> which I think ought to be xml:base. There's similar problems for the
> dfxp-transformation profile, though dfxp-full looks OK.
>

I have researched these discrepancies and determined the following:

(1) the standalone profiles referenced in Shaun McCance's original message
appear to be modified snapshots of the TTML1 First Edition (1E) profiles
specified in the interval of May through September of 2009 [1][2][3];

[1] http://www.w3.org/TR/2009/WD-ttaf1-dfxp-20090511/
[2] http://www.w3.org/TR/2009/WD-ttaf1-dfxp-20090602/
[3] http://www.w3.org/TR/2009/CR-ttaf1-dfxp-20090924/

I say this because because:

   - the May 2009 WD used the base attribute instead of xml:base, which was
   changed to xml:base by the June 2009 WD;
   - the May 2009 WD did not define a dfxp-full profile, while the June
   2009 WD did define it;
   - the TTML Feature and Extension namespaces in the June 2009 WD did not
   have a final '/', while the September 2009 CR did have the final '/';

(2) the current standalone profiles (prior to any update we will make in
the upcoming days) appear to be slightly different from what Shaun
describes, because they do use the xml:base attribute, and they do specify
the terminating '/' on the TTML Feature and Extension namespaces;

(3) additionally, the current standalone profiles (prior to any update we
will make in the upcoming days) differ from the currently defined TTML1
Second Edition (2E) profiles as follows:

   - the dfxp-presentation profile specifies as required the following
   features which corresponds to the May 2009 [1] spec, changed to optional in
   June 2009 [2]:
      - #color
      - #display-block
      - #display-inline
      - #extent
      - #fontSize-isomorphic
      - #fontStyle-italic
      - #fontWeight-bold
      - #length-cell
      - #length-pixel
      - #length-positive
      - #styling-inheritance-content
      - #styling-inline
      - #textAlign-absolute
      - #visibility-block
      - #visibility-inline
   - the dfxp-transformation profile specifies as optional the following
   features which corresponds to the May 2009 [1] spec, changed to required in
   June 2009 [2]:
      - #content
      - #time-offset
      - #timing
   - the dfxp-full profile specifies as required the following features
   which corresponds to the May 2009 [1] spec, but removed from the profile in
   the TTML10 1E PR in September 2010 [4]:
      - #dynamicFlow-character
      - #dynamicFlow-clear
      - #dynamicFlow-fill
      - #dynamicFlow-glyph
      - #dynamicFlow-in
      - #dynamicFlow-jump
      - #dynamicFlow-line
      - #dynamicFlow-out
      - #dynamicFlow-rollUp
      - #dynamicFlow-smooth
      - #dynamicFlow-teletext
      - #dynamicFlow-word
      - #dynamicFlow
      - #rollUp
   - the dfxp-full profile specifies as required the following feature
   which corresponds to the May 2009 [1] spec, but removed from the profile(s)
   in in September 2009 [3]:
      - #fontSize-anisomorphic
      - #smpteMode
      - #textOutline-blur
   - the dfxp-full profile specifies as required the following feature
   which corresponds to the Map 2009 [1] spec, but removed (as a dangling
   reference) from the profile(s) in TTML10 2E in September 2013 [5]:
   - #overflow-scroll
   - the dfxp-full profile does not include any reference to the following
   feature which corresponds to the May 2009 [1] spec, but added to the
   profile in the CR of September 2009 [3]:
      - #fontSize-anamorphic (renamed from #fontSize-anisomorphic)
      - #overflow-visible
      - #textOutline-blurred (renamed from #textOutline-blur)
      - #textOutline-unblurred
   - the dfxp-full profile does not include any reference to the following
   features which corresponds to the May 2009 [1] spec, but added to the
   profile in the TTML1 1E PR in September 2010 [4]:
      - #clockMode-gps
      - #clockMode-local
      - #clockMode-utc
      - #dropMode-dropNTSC
      - #dropMode-dropPAL
      - #dropMode-nonDrop
      - #dropMode
      - #extent-region
      - #extent-root
      - #fontStyle-oblique
      - #length-integer
      - #markerMode-continuous
      - #markerMode-discontinuous

[4] http://www.w3.org/TR/2010/PR-ttaf1-dfxp-20100914/
[5] http://www.w3.org/TR/2013/REC-ttml1-20130924/

We now need to update the standalone profiles to match the content of TTML1
2E. It should be noted that, with one exception, all the differences
described above correspond to changes made to TTML1 prior to publishing of
1E as REC in September 2010. Given that only the final TTML1 1E spec
profile definitions applied as of September 2010, the differences in the
standalone profile documents should be considered as errors, and the final
1E REC specification text given priority. If there are interoperability
issues that arise from those differences, then the implementations and/or
content should be changed accordingly, since it would not adhere to TTML1
1E.

The one exception I mention above is that we removed in TTML1 2E the
dangling reference to #overflow-scroll that appeared in the three TTML1 1E
profiles. That is, the three profile documents in TTML1 1E contained a
reference to a previously removed feature definition #overflow-scroll,
which was effectively undefined in 1E.

As far as implementations of the dfxp-presentation profile which chose to
follow the standalone profiles rather than the published 1E spec, they
would have implemented 15 features that they didn't need to in order to
satisfied the published spec. Content that relied on such an implementation
may fail to operate on a correct implementation of 1E which did not require
these 15 features.

For dfxp-transformation, the published 1E spec added three required
features not specified in the standalone profile (based on the earlier
draft), so content that relied on such an implementation would continue to
operate on a correct implementation of 1E which requires these additional
features.

For dfxp-full, it is more of a mixed picture, as the standalone profile
(based on a combination of pre-1E and final 1E versions) both requires more
than, but doesn't require all of, what is required in the final 1E. Content
that relied on such an implementation may or may not continue to operate on
a correct implementation of 1E depending on which features are used.

My overall conclusion, is that we should update the standalone dfxp-*
profiles immediately to the published 2E profile definitions, and let the
chips fall as they may.

G.