RFC2518 bis, attributes on property names -- in or out?

I heard from at least one implementor (Joel Soderberg) that random
attributes on property names should not be stored "as part of" the
property.  This makes it more complicated to marshall the property again
when responding to a PROPFIND request.  Attribute namespaces may be
difficult here. What namespace are the attributes in?  What if the
attribute on the property name is a namespace declaration -- does that
have to be preserved *in that location*?

Another consideration is that we may want to reserve attributes on the
property name for meta-information *about* the property (like data
types).  That kind of information isn't part of the *value* of the
property, but it may be used to figure out how to treat the property now
or later.  An implementation could safely ignore attributes it didn't
understand.  

RFC2518 says "Language tagging information in the property's value (in
the "xml:lang" attribute, if present) MUST be persistently stored along
with the property, and MUST be subsequently retrievable using PROPFIND."
I believe the spec is silent about other attributes.

In RFC2518bis, we got more specific about what makes up the value of a
property:
   "The value of a property consists of attributes on the property name 
   element, language attributes which are scoped to the property, 
   namespaces which are used in the property name element or its 
   children, and child elements including text.  The server MUST 
   persistently store this information and reconstruct it in PROPFIND 
   responses.   "

Now that objections have been raised, who feels strongly for one side or
another?

Lisa

Received on Wednesday, 27 November 2002 21:38:23 UTC