[WF-Protocols] first attempt with xml schema

Matze Braun matze at braunis.de
Mon Jul 7 23:37:59 PDT 2003



On Mon, 7 Jul 2003, Anders Stenberg wrote:

> res wrote:
> 
> > On 06.07.2003 21:09, Matze Braun wrote:
> >
> >> -How should we reference to the textureunits? currently you can give 
> >> a integer
> >>   number like this:
> >>   <texturecoords textureunit="1"/>
> >
> >
> > IMHO "texture units" sound too hardware-centric. Maybe "texture 
> > coordinate set" is a better name. Also, identifiers for set should be 
> > arbitrary names:
> >     <texturecoords set="diffuse_map" />
> >     <texturecoords set="fancy_effect" />
> > etc.
> >
> > - f.r.
> >
> I guess I agree here. I'd like to see some kind of general entry too. 
Ok, I changed this to a "set" attribute with string identifier.

> Like <customcomponent name="tangent"><c x="0" y="1" z="3.4">.... I think 
> otherwise it'd be a bit constrained, and people will probably start 
> storing custom data (like tangents, binormals, special color 
> assignments, material blend factors, whatever) in places where it's not 
> supposed to be, just to get it in there somehow. (For example use some 
> texturecoordinate set that isn't used, or something.) Do I make sense here?

As noted in some earlier mails we should NOT support general entries. 
Anyway xml namespaces already provide us with a good way to deal with such 
issues. You can simply put your special stuff in a separate namespace:
<mysp:strangetangents>
  <mysp:c x=".." y=".." z=".."/>
  <!-- ... -->
</mysp:strangetangents>

So far from the technical side. From the practical side we should try to 
avoid custom extensions at all costs. The results will be that some 
converters support more features than others, so information is lost 
between conversion runs and the standard will get very weak. The best 
solution for customisations is thinking about all important usecases while 
designing the format. There should be no need to add custom extensions for 
users/developers. If something is missing then we should try getting into 
the next version of the spec before the extensions become a 
"semi-standard".

> 
> Also, I'm a bit split about naming the positions "vertices", like from 
> Matze's example:
> 
>     <vertices>
>       <v x="-1" y="-1" z="-1"/>
>       <v x="1" y="-1" z="-1"/>
>       <v x="-1" y="1" z="-1"/>
>       <!-- More vertices come here. -->
>     </vertices>
> 
> I see "vertex" as the whole collection of position, normal, texture coordinate, etc. For me "position" or something would be a bit more appropriate. Maybe that's just me. It's not a really important issue to me, just an opinion. :)
Hmm true. What about the name "point"?

Greetings,
	Matze



More information about the Protocols mailing list