RE: [wsbpel] RE: NCName restriction to avoid "."

[Thomas Schulze <ThomasSchulze@de.ibm.com>]

+1 to the idea catching this by the schema validation.

Isn't this constrain only needed for those places where new variables are
declared? That should be <variable name="...">, <forEach
counterName="...">, <onEvent variable="..."> and the <fromPart
toVariable="..."> version of <onEvent> and <catch faultVariable="...">.

All other cases are variable references. Not sure if we really need it
there.

Btw. doesn't the <fromPart toVariable="..."> variant of <onEvent> require
the variable's type defined (attributes messageType, element and type)?

Best regards/Mit freundlichen Grüßen,

       Thomas Schulze



                                                                       
             "Mark Ford"                                               
             <mark.ford@active                                         
             -endpoints.com>                                            To
                                       "'Alex Yiu'" <alex.yiu@oracle.com>
             05.06.2006 01:35                                           cc
                                       "'wsbpeltc'"                    
                                       <wsbpel@lists.oasis-open.org>   
                                                                   Subject
                                       RE: [wsbpel] RE: NCName restriction
                                       to avoid "."                    
                                                                       
                                                                       
                                                                       
                                                                       
                                                                       
                                                                       




Yes. Basically, anywhere that we're referring to a variable the type should
be "BPELVariableName" instead of "NCName". Of course I'll leave the actual
selection of the name to the schema powers at be but you get the idea.

The only downside to this is that we need to add a little more text to the
current spec to define what this type is since it will appear in our pseudo
schema examples. This may impact readability a little but it goes a long
way in making the restriction explicit. If I were reading the spec for the
first time and I saw "BPELVariableName" for the data type I would naturally
assume that there was some restriction in play and then go and find the
definition of this type. As it is now, NCName appears in multiple places
without any kind of qualifier to indicate that it's a restricted NCName.

From: Alex Yiu [mailto:alex.yiu@oracle.com]
Sent: Sunday, June 04, 2006 7:26 PM
To: Mark Ford
Cc: 'wsbpeltc'; Alex Yiu
Subject: Re: [wsbpel] RE: NCName restriction to avoid "."


Mark,

Do we want to apply the same restriction to variable attribute of <from>
and <to> spec also?

Thanks!

Regards,
Alex Yiu


Mark Ford wrote:
      +1 from me ;)

      This would also apply to the implicit variables created by a
      <fromPart> nested within an onEvent.

      I'll open an issue on this.

      From: Alex Yiu [mailto:alex.yiu@oracle.com]
      Sent: Sunday, June 04, 2006 5:38 PM
      To: wsbpeltc
      Cc: Mark Ford; Alex Yiu
      Subject: NCName restriction to avoid "." (was: [Fwd: BPEL Schema
      question])


      Hi guys,

      Mark has the following suggestion to use the data type other than
      NCName to restrict the "." usage in variable names.

      I am OK with it.
      Thought?

      Thanks!

      Regards,
      Alex Yiu


      -------- Original Message --------
                                                                       
     Subject: BPEL Schema question                                     
                                                                       
        Date: Sun, 4 Jun 2006 13:27:35 -0400                           
                                                                       
        From: Mark Ford <mark.ford@active-endpoints.com>               
                                                                       
          To: 'Alex Yiu' <alex.yiu@oracle.com>                         
                                                                       





      Hi Alex,


      What do you think about changing the data type of a variable's name
      to be a restriction of NCName to disallow the "." as described in
      Section 8.1? The dot is a legal value for NCNames but disallowed for
      Xpath binding purposes of message variables.


      This also applies to the name of the counter attribute on a
      <forEach>.


      I'd prefer to catch these types of errors with the XSD as opposed to
      having to write code to catch them during static analysis.


      Thanks.