RE: Issue 302: Graph edges that do not terminate

So if we mandated xsi:nil, the array case would be fine?
 
The only case that I know of where we have an 'inbound' only edge is the first edge in the graph ( the one represented by the top-level element of the serialization ). One could argue that this is not really an edge, but only a node. I'd be happy to amend the doc along those lines if that's what people want.
 
Gudge

 -----Original Message----- 
 From: Jacek Kopecky [mailto:jacek@systinet.com] 
 Sent: Mon 9/9/2002 12:07 
 To: Martin Gudgin 
 Cc: XMLP Dist App 
 Subject: Re: Issue 302: Graph edges that do not terminate
 
 

  Gudge, others,
  the text below has a big problem with arrays because it doesn't allow
 arrays with nils elsewhere than at the end - that's because the edges
 are identified by position and if an edge (that could have been there in
 a different situation) is not there, well, that changes the positions of
 the edges after it.
  Now the text in the editors' copy introduces inbound-only edges that
 are not (AFAICS) serializable using the SOAP Encoding rules. I think
 these should be removed or dealt with in the SOAP Encoding. I don't
 really think this would be merely an editorial change if serialization
 for inbound-only edges was added to SOAP Encoding.
  So while at first I liked the alternative below better, I now think
 that the text from the editors' copy is the right way to go, although it
 does need some more attention.
  Best regards
 
                    Jacek Kopecky
 
                    Senior Architect, Systinet Corporation
                    http://www.systinet.com/
 
 
 
 On Mon, 2002-09-09 at 00:08, Martin Gudgin wrote:
 >
 > I took an action item at last weeks concall to propose resolution text
 > for Issue 302[1]
 >
 > It turns out I had already incorporated such text into the editor's copy
 > of part 2[2] as part of the resolution to Issue 353[3] ( classified
 > editorial ).
 > The text can be found in green highlight at[4]. Also refer to clause 4
 > of[5].
 >
 > If people are unhappy with the resolution, perhaps the following would
 > be preferable:
 >
 > 1. Remove the green highlighted text from[4]
 >
 > 2. Amend clause 4 of[5] to read:
 >
 >       Certain graphs may sometimes contain a given edge and at other
 > times that edge will be missing. Such missing edges can either be
 > omitted from the      serialization or can be encoded as an element
 > information item with an xsi:nil attribute information item whose value
 > is "true".
 >
 > On the whole, I think I prefer the above, rather than what is in the
 > editor's copy. The problem with the editor's copy is that there is no
 > way to determine the label of an edge which does not terminate in a
 > graph node. While this is OK for an outbound edge of an array, it is not
 > OK for an outbound edge of a struct. The above resolution draws out the
 > fact that the edges were not present in the graph at serialization time.
 >
 > If we mandated xsi:nil then there would be an edge label, so the above
 > concern would go away, but I'm not sure anyone want's to go there right
 > now.
 >
 > Comments, flames, discussion etc. to the usual address.
 >
 > Gudge
 >
 > [1] http://www.w3.org/2000/xp/Group/xmlp-lc-issues#x302
 > [2] http://www.w3.org/2000/xp/Group/2/06/LC/soap12-part2.xml
 > [3] http://www.w3.org/2000/xp/Group/xmlp-lc-issues#x353
 > [4] http://www.w3.org/2000/xp/Group/2/06/LC/soap12-part2.xml#graphedges
 > [5] http://www.w3.org/2000/xp/Group/2/06/LC/soap12-part2.xml#complexenc

Received on Monday, 9 September 2002 09:54:25 UTC