Re: [DotGNU]Dgmx and the metadata problem

[Gopal V]

If memory serves me right, Peter Minten wrote:
> For those languages you could simply precompile the metadata into dgmx 
> and load it at runtime. 

That's the idea allright ...

> So when I would write the state example in Ruby it would look like this:
> 
> #[WebService(
> # Description="Returns State name (from example in CookComputing.XmlRpc)", \
> # Namespace="http://nfluid.com";)]
> class StateName
>   #[WebMethod("examples.getStateName", \
>   # Description="Returns the name of a state given its number")
>   # Args={int}]
>   # Note that since Ruby doesn't have typed arguments the Args attribute arg
>   # tells the metadata compiler what to write as argument types.

I would like the Python module to re-use the __doc__ thingy like this ...

def MyMethod(x):
        """
                [WebMethod("examples.getStateName",
                        Description="Returns the name of a state given its 
number",
                        Args="String;(int;)"]
        """
        return state[x];

Which would make it simpler for python people to just pull out the
MyMethod.__doc__ and XmlRpc being our main RPC model right now , singleton
methods do have an advantage over instance methods in classes ....

I wonder how we're going to expose the same method via SOAP and XMLRPC ...

For DGEE to be perfect in some sort of way we need to finish the

        public virtual void GetObjectData (SerializationInfo info, 
                                                                                
StreamingContext context)

for *each* serializable class. And call them with an XmlRpcSerializationInfo 
and A DGEEStreamingContext. It might be easy , but we have to take care not to
lock out other .NET implementations people might use to interact with DotGNU. 

Gopal
-- 
The difference between insanity and genius is measured by success