RE: [cgmo-webcgm] ISSUE: In what components is 'transform' defined?

["Galt, Stuart A" <stuart.a.galt@boeing.com>]

Dieter,

> DW: So you are talking about a CGM containing regular 
> grobjects only that should be considered as segments at runtime?

We do not allow access at the element level today, everything needs to
be rolled up to the APS container.  A hard line approach would be only 
APS structures that contain a BEGIN SEGMENT or COPY SEGMENT construct 
within the bounds of the APS could be considered segments at runtime but
this has
a couple problems... a) It doesn't handle the case where there is
geometry
other than the segment contained within the APS.  b)It wont allow a
version
1.0 or 2.0 file to be used.

a) can be handled by making a rule that all of the geometry is affected
by
the transform (similar to how the rest of the APS level changes work)
and b) 
can be handled by making it possible to create a transform via something

other than requiring a segment to exist in the CGM file.

The reason that I brought up segments was that it is a mechanism that 
exists in CGM but on WebCGM and it defines how to create static
transforms.
I realize that it might be not very well supported by the viewers or 
the editors.  Does anyone support segments or is it one of those dark 
corners of CGM that nobody has implemented?

Even if we don't use segments, I think that the transforms we have been
discussing and the CGM segment are similar constructs.  To finally 
answer your question... I guess I would treat all APS as just "normal"
structures until I was forced to treat it as a transformed (or segment)
structure.  I see four potential ways of doing this:
1) DOM call to transform an APS
2) XCF binding a transform to an APS
3) CGM metafile contains a attribute (or style property) that associates
   a transform to an APS
4) The APS contains only a segment (or transform + copy segment)

Once one of the above conditions has been met the APS needs to be
treated as a transformed APS or segment or whatever we end up calling
it.
The first three have been part of the proposal for quite some time.
Allowing segments is what makes the fourth option possible.



> DW: sure, this is straightforward. Since the style 
> properties/transform attributes are applied using DOM or XCF 
> they will work as well as setting a different lineweight or 
> adding a namespaced attribute to a WebCGM 1.0 file.

Yes, I agree.


Stuart.
--
Stuart Galt
SGML Resource Group
stuart.a.galt@boeing.com
(206) 544-3656

 

> -----Original Message-----
> From: Weidenbrueck, Dieter [mailto:dweidenbrueck@ptc.com] 
> Sent: Friday, January 18, 2008 10:52 AM
> To: Galt, Stuart A; Cruikshank, David W; Lofton Henderson; 
> cgmo-webcgm@lists.oasis-open.org
> Subject: RE: [cgmo-webcgm] ISSUE: In what components is 
> 'transform' defined?
> 
> Stuart,
> 
> again, inline... 
> 
> -----Original Message-----
> From: Galt, Stuart A [mailto:stuart.a.galt@boeing.com]
> Sent: Freitag, 18. Januar 2008 19:30
> To: Weidenbrueck, Dieter; Cruikshank, David W; Lofton 
> Henderson; cgmo-webcgm@lists.oasis-open.org
> Subject: RE: [cgmo-webcgm] ISSUE: In what components is 'transform'
> defined?
> 
> My responses are below.
> 
> --
> Stuart Galt
> SGML Resource Group
> stuart.a.galt@boeing.com
> (206) 544-3656
> 
>  
> 
> > -----Original Message-----
> > From: Weidenbrueck, Dieter [mailto:dweidenbrueck@ptc.com]
> > Sent: Friday, January 18, 2008 9:43 AM
> > To: Galt, Stuart A; Cruikshank, David W; Lofton Henderson; 
> > cgmo-webcgm@lists.oasis-open.org
> > Subject: RE: [cgmo-webcgm] ISSUE: In what components is 'transform' 
> > defined?
> > 
> > Stuart,
> > 
> > interesting idea about segments, however, that would 
> require to turn 
> > every addressable object into a segment. So there is a significant 
> > overhead just for the purpose of somebody eventually applying a 
> > transform.
> 
> Would it be possible to only create a segement at the 
> application structure level if it is actually used in the 
> metafile or called out in the xcf?  I was just trying to see 
> if we could use or extend existing constructs.  I agree 
> turning every APS into a virtual segment would cause too much 
> overhead.
> 
> > 
> > It would also mean that transforms could not be applied to 
> WebCGM 1.0 
> > or 2.0 (and hence ATA and S1000D) files, which is not really 
> > acceptable.
> 
> I am not sure how we could do something in 2.1 that would 
> suddenly make transforms possible with 1.0/2.0 data.  Or were 
> you suggesting that I could take a 2.1 XCF file and a older 
> CGM file and have it work with a
> 2.1 viewer?
> DW: sure, this is straightforward. Since the style 
> properties/transform attributes are applied using DOM or XCF 
> they will work as well as setting a different lineweight or 
> adding a namespaced attribute to a WebCGM 1.0 file.
>  
> > The math is clear, and we already know that we need 3x3 
> matrices (to 
> > be able to include shearing). That should make our 
> discussion an easy 
> > one, I guess. This is what SVG does, this is what PostScript does.
> 
> Yes the 3x3 matrix is what is needed and is used by most 
> things that do a 2D to 2D coordinate transfomation.  Only six 
> terms in the matrix need to be specified because the bottom 
> row is not really needed to produce the new 2D coordinate. 
> 
>