RE: [cgmo-webcgm] 1.0 tests modified for 2.0

[Lofton Henderson <lofton@rockynet.com>]

At 08:46 AM 2/13/2006 +0100, Dieter  Weidenbrück wrote:
> > -----Original Message-----
> > From: Lofton Henderson [mailto:lofton@rockynet.com]
> > Sent: Sunday, February 12, 2006 11:09 PM
> > To: cgmo-webcgm@lists.oasis-open.org
> > Subject: [cgmo-webcgm] 1.0 tests modified for 2.0
> >
> > Information, and QUESTION/ISSUE (for email and for next telecon)...
> >
> > On FTS:  webcgm20-ts-20060212.zip
> >
> > I have started modifying the 1.0 (rel-1.1) dynamic tests, for
> > 2.0 correctness, according to the assessment I sent a while
> > ago.  I will keep doing this, and periodically put up new zip files.
> >
> > In this batch, changed tests are:
> > * linking-selectID-BE-05
> > * linking-selectName-BE-06
> > * linking-anyURI-BE-07
> >
> > You can view them by opening the IntroPage.html and
> > navigating from there.
> >
> > These three tests have something in common:  they have
> > navigation (links) to objects, but no specified object
> > behaviors.  Therefore they use the default object behavior.
> > We changed the default object behavior for 2.0.  In 1.0, it
> > was effectively zoom+newHighlight (with a small wrinkle about
> > presence/absence of a 'viewcontext' ApsAttr on the target).
> >
> > Therefore you will see this:
> >
> > 1.) a 1.0 viewer showing the 1.0 file (in the 1.0 Test Suite)
> > should give unzoomed view, highlighted object.
>yes
> >
> > 2.) a 2.0 viewer showing the 2.0 file should give zoomed
> > view, highlighted object.
>yes
> >
> > 3.) QUESTION.  what about a 2.0 viewer on the 1.0 file?
> > Should it detect the version of the CGM (target) and do #1 or
> > #2 accordingly?
>We defined a mapping for 2.0 viewers for 1.0 behaviors. Once the
>behaviors got mapped, the viewer needs to show the correct behavior
>for the 2.0 behaviors.

Yes, the 1.0 default is 'view_context', which 2.0 (CS) says to map to 
zoom+newHighlight, regardless of the attributes of the target object.  I'm 
going to question whether we goofed here...

The 1.0 'view_context' object behavior is this:
a.) zoom+newHighlight if there is a 'viewcontext' ApsAttr on the target object;
b.) full+newHighlight if not.

I would be tempted to argue that, in real-world 1.0 content, there will be 
many more default cases of type #b (no 'viewcontext' ApsAttr present) than 
type #a, which is the single default we chose for 2.0.   From a standpoint 
of user-friendliness, it would seem better not to introduce a dramatic 
viewing change to what I guess to be the preponderance of WebCGM 1.0 
content in the real world.

Reinforcing this, the default cases in the 1.0 test suite don't have 
'viewcontext' ApsAttr present.  I'm questioning the wisdom of a 
specification that is not actually necessary and that forces a change of 
behavior by all conforming 2.0 viewers on virtually all default cases in 
the test suite, and (IMO) most default cases in the real world.

Remind me, why did we not simply copy this behavior (a plus b) for the 2.0 
default?  All 1.0 viewers could do it, with no changes.  It is not as 
simple to write down as a single value ...  but it's already implemented!.

It was not until I started working on the test suite that the 2.0 default 
began to bother me (this is the value of a TS, and it would have been nice 
if we had had the resources to upgrade it concurrently with the 
spec).  About 1/4 of the tests are affected.

Bottom line -- I can't imagine 1.0 users being very happy, when their 
vendor upgrades their viewer to 2.0, that all of their 1.0 content starts 
behaving differently.

Regards,
-Lofton.