Worry about using lexical order for preferences.

["Moberg, Dale" <Dale_Moberg@stercomm.com>]

Hi

One idea under consideration in TPA WG is 
to use the CPP lexical order for preferences
among those elements having duplicated 
capabilities (such as several possible
delivery channels and so on).

I noticed yesterday that a popular 
implementation of DOM API in Java (Xerces)
has a number of methods returning 
multiple elements that do not preserve
lexical order. For example, the "getAttributes" 
method returns a NamedNodeMap
in which the nodes are alphabetically 
ordered, so that index 0, for example,
returns the alphabetically first 
attribute by name of attribute and not the
lexically first attribute (I guess 
I'd better mention that lexical order here
just means the order in which 
it occurs in the file/stream).

So while I have not checked out every 
method (and I think this will be a DOM
problem, because SAX callbacks do 
follow lexical order as far as I can tell),
I am wondering whether we should 
be more explicit about preference. Sorry
to add another item. 

(I should also mention that
DOM implementations will probably
be popular for dealing with CPP
processing because of the need to jump around
with IDs while figuring out whether
and how the capabilities match.)

Dale Moberg