Re: [cgmo-webcgm] ISSUE: is apsid an attribute?

[Benoit Bezaire <benoit@itedo.com>]

Hi Lofton,

See inline...

Thursday, July 7, 2005, 9:20:30 PM, Lofton wrote:

LH> Source:  editors / test writers.

LH> ISSUE:  does the apsid parameter of the BegAps element show up as an
LH> attribute in the DOM?

LH> DESCRIPTION:

LH> WebCGMNode contains a boolean method, hasAttributes(), and an attribute of
LH> type node-list, 'attributes'.  What qualifies as an attribute is not
LH> defined anywhere.
Right and we should probably correct that.

LH>   The obvious thing that comes to mind is APS Attributes
LH> (on APS nodes).  The text says the value of 'attributes' will be null in
LH> the case that the WebCGMNode has no attributes.

LH> WebCGMAppStructure, which inherits from WebCGMNode, has an additional
LH> attribute, 'apsId'.
Right, just like the DOM/SVG element interface has an 'id' attribute.
It's a convenience method that is frequently used by script writers.

LH>   In metafiles, the apsid is a parameter on the BegAps
LH> element (along with the 'type' and the 'inheritance flag' parameters).
Yes, the DOM and CGM meaning of apsid are not exactly the same.

LH> The initial code of the test Node-attr.html shows that the apsid is being
LH> treated in the test as if it is an 'attribute'.
I did the test and that's how our implementation currently works.

LH> So should DOM return the metafile apsid as an 'attribute' (WebCGMNode
LH> interface), or only as the 'apsId' on the WebCGMAppStructure interface, or
LH> both?
I went for both, it would be good if people could share their thoughts
on this. I thought we agreed on that, but it probably wasn't a minuted.

LH> It seems to me that the way we designed the interfaces and their 
LH> methods/attributes seems to hint at "only 'apsId'".  On the other hand,
LH> Benoit seems to remember some past decision for 'attribute' or "both".  If
LH> that is the case, note that hasAttributes() will always return true for an
LH> APS node (by far its most usual usage, if not the only usage), and
LH> 'attributes' will always have a count of at least 1.  Also note that in
LH> XCF, apsid is a required XML attribute on all the APS elements (and the
LH> bindById, of course).

LH> It is not really critical *how* we answer it.  We have some freedom in how
LH> we map the structured metafile instance onto an XML-like DOM tree.  But it
LH> is critical that we have a clear answer and document it.
Yes.

I don't have a really strong opinion about this... but in XML, 'id' is
an attribute; and we have apsid required on all XCF elements (except
for bindByName). We have ask ourselves if we want a DOM that script
writers will easily understand or if we want a DOM that is very closely
linked to the CGM format (which only this group understands).

LH> ALTERNATIVES:
LH> Alt.1:  'apsId' only
LH> Alt.2:  'attribute' only (which would mean eliminate 'apsId')
LH> Alt.3:  both

LH> RECOMMENDATION:  none yet, TC should discuss.  Implementors, what have you
LH> done?
I don't have a strong opinion, I'll vote for Alt 3 mostly because I
don't feel like changing my code; but also because it is harmless.

LH> Regards,
LH> -Lofton.

-- 
 Benoit   mailto:benoit@itedo.com