W3C home > Mailing lists > Public > www-xml-schema-comments@w3.org > October to December 2000

An observation on XML schema limitation

From: ALEXANDRU,MARIUS (HP-Romania,ex1) <marius_alexandru@hp.com>
Date: Fri, 6 Oct 2000 16:27:57 +0200
Message-ID: <C7773A81764BD4118A2100D0B747AC377AEF62@tell.gva.hp.com>
To: fallside@us.ibm.com
Cc: www-xml-schema-comments@w3.org

   Hi,
     I apologize for this mail but I want to present you a strong limitation
of Schema model comparing with the entity-relationship model that maybe
could be considered at your meetings regarding the xml-schema draft.
   
     Suppose that you have to entities :
        
                     <entity name="a" id="a">
                        ...
                     </entity>

                     <entity name="b" id="b">
                        ...
                     </entity>

      and a relationship Rab(a,b) between the  two entities :
                     <relation name="Rab">
                         <role name="has-a" occurs="0,n"/>    
                            <entity idref="a"/>   
                         </role>
                         <role name="has-b" occurs="0,n"/>    
                            <entity idref="b"/>   
                         </role>
                      </relation>

     In my opinion, the E-R model (I don't used an exact formalism to
describe it in the above example) can allow us to produce the following 
valid XML productions:

   Valid Production 1:   
            <a> 
              <Rab>   
                 <!-- we have to include also the rolename here too-->
                 <b>
                  ...
                 </b> 
                 <b>
                  ...
                 </b> 
                 <b>
                  ...
                 </b>     
              </Rab>
            </a>

   Valid Production 2:   

            <b> 
              <Rab> 
                 <!-- we have to include also the rolename here too-->
                 <a>
                  ...
                 </a> 
                 <a>
                  ...
                 </a> 
                 <a>
                  ...
                 </a>     
              </Rab>
            </b>

  Valid Production 3:   
           <Rab>
               <!-- we have to include also the rolename -->
              <a>
                ...
              </a>
              <b>
       	  ...
       
              </b>
           </Rab>  
            <!-- or another description that allow us to introduce more
couples on a single <Rab>...</Rab> chunk.
           
   So as you can see the XML Schema definition (at www.w3.org) don't allow
us to produce all the productions that the E-R model could provide.
   In fact the problem is more important that this : the tree (xml)
structure its in itself a very constraintfull syntax (so a lot of patterns
have to be specified (like the already existing :
sequence,conjuction,disjunction etc..) in order to give more flexibility of
this strong syntax.
   Also, I cannot imagine a good XML-Query definition that will depend so
strongly on the Syntax in which the data are presented.

   I also think that this limitation will affect the IT industry for a long
period of time since now (please take this last sentence not so serious as
it seems, but in any case is my opinion)

     Thanks for your time and Best regards,
                    Marius
 
Received on Friday, 6 October 2000 10:28:16 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 6 December 2009 18:12:48 GMT