[LON-CAPA-dev] RFC: XML for <imageresponse> rewrite

[Guy Albertelli II]

Hi Gerd,
> >
> >   <vector>
> >      <angle unit="degrees">
> >        35
> >        <responseparam name="tolerance" type="tolerance"  
> > default="3%" />
> >      </angle>
> >      <length unit="pixel">
> >        100
> >        <responseparam name="tolerance" type="tolerance"  
> > default="3%" />
> >      </length>
> >   </vector>
> >
> 
> Looks weird, somehow. Also, not sure why on that level. My earlier  
> proposal was
> 
> <vector>
>       <angle unit="degrees">35</angle>
>       <length unit="pixel">100</length>
>        <responseparam name="tolerance" type="tolerance" default="5" />
>    </vector>
> 
> where it is a 5 pixel radius around the end point. I would not quite  
> know how to visualize relative errors on vectors - why would I have  
> more tolerance around a 60 degree angle than around a 10 degree angle?

Well my thinking is you might want different tolerances on the angle
versus the lenght.

I honestly don't know what you example means when I see it. (5 what?
pixels or degrees? What if I want 0.1 radians and 10 pixel tolerance?)

Which is why I'll need to make the tol apply to the number, of the
object, of the answer, of the response, of the part.

And the same for this eaxmple: 

> >
> > - to address C), <vector> gets a 'style' (cartesian, or polar) and
> >
> >   <vector style="cartesian">
> >      <xlength unit="pixels">
> >        35
> >        <responseparam name="tolerance" type="tolerance"  
> > default="3%" />
> >      </xlength>
> >      <ylength unit="percentage">
> >        10
> >        <responseparam name="tolerance" type="tolerance"  
> > default="3%" />
> >      </ylength>
> >   </vector>
> 
> Okay. Again, I cannot quite see when I would need two tolerances, and  
> it does look odd to have the tolerances folded in like that.
> 
> <numericalresponse unit="N/m" answer="15">
> <responseparam name="tolerance" type="tolerance" default="3%" />
> </numericalresponse>
> 
> seems to follow a different philosophy.

Because it right now only (in most usage) accepts a single number.
-- 
guy@albertelli.com   0-7-2-1-27,137