perfsonar development work

[Jason Zurawski <>]

maxim wrote:
> Hi Yee,
> I am still working on your schema. One thing is clear - it should be
> independed to current ping.rnc
> since extending it will create more mess. 

I would agree that it should be separate as well.

> Another this is that even without
> dependence the schema is not valid
> due some inherited problems from underlying schemas. They all written in
> Relax NG which makes them more expressive
> but less valid when translated into the XML schema world. For example ( and
> this is a question to the person
> who designed it, its nmtopo-l4.rnc file ) this fragment is not valid (
> elements with the same name must be of the same type):
>   

The tooling I regularly use claims warning on this, which I have always 
interpreted as being usable for our [non-validated] purposes.  We could 
of course re-write the various fragments so the validator is happy, but 
we loose a little bit of power (i.e. a single 'endPoint' with roles that 
can be 'src' or 'dst' can lead us to something that has two of either). 

>  L4EndpointPairContent =
>         (
>                 element nmtl4:endPoint {
>                         attribute port { xsd:string }? &
>                         attribute protocol { xsd:string }? &
>                         (
>                                 attribute role { "src" } |
>                                 element nmwgtopo3:role { "src" }
>                         )? &
>                         (
>                                 element nmtl4:address { L4Address } |
>                                 element nmtl3:interface { anyThing }
>                         )?                
>                 },
>                 element nmtl4:endPoint {
>                         attribute port { xsd:string }? &
>                         attribute protocol { xsd:string }? &
>                         (
>                                 attribute role { "dst" } |
>                                 element nmwgtopo3:role { "dst" }
>                         )? &
>                         (
>                                 element nmtl4:address { L4Address } |
>                                 element nmtl3:interface { anyThing }
>                         )                
>                 } 
>         )
>
> Can someone let me know how to deal with that ?
> Also in the nmbase.rnc this will violate UPA principal :
>   

The usual problem that will violate this is the choice between attribute 
and element:

<element name="foo" value="bar"> vs. <element name="foo">bar</element>

The 'anyElement' trick also probably allows for an ambigous parse (i.e. 
an element like <subject> can enter into it's specified pattern, or be 
picked up by 'anyElement'.

Because we have never forced strict validation at the code level this 
hasn't been an issue, but if anyone has suggestions to tighten this up 
it would be great.  In my opinion it is important to have 'legal' XML 
more than completely validated XML, the individual implementations know 
exactly what they need and do a great deal of the rejection. 

-jason