Re: Service description value for hypermedia?

On Jun 16, 2005, at 8:56 PM, Marc Hadley wrote:

> On Jun 16, 2005, at 2:41 PM, Mark Baker wrote:
>
>>
>> But the code that a human would write to support Atom as a library
>> (e.g. http://www.howdev.com/technologies/), would already provide an
>> API analogous to the one you describe there.
>>
>>
> Indeed, Atom is a bad example since its likely that folks will want  
> to write custom libraries for most programming languages to support  
> the format and protocol. Is that the case for all web applications  
> though...
>

I disagree. Whether folks write custom libraries or not is beside the  
point, IMO.

Code generation makes sense if one can extract a pattern that all of  
those libraries could have in common. If there is such a thing, it's  
useful to separate the pattern from the specifics, describe the  
specifics with the description language, and put the implementation  
into the code generator.

When we do code generation in our MDA projects, we examine a few use  
case implementations and look for similarities. If we find them,  
refactoring them into code generation templates is one of many  
options. The same might be true here: If convenient libraries for  
Amazon, Atom, Google and whatever-RESTful-app share a lot of code  
patterns that are not easily or efficiently put into a framework,  
they might be candidates for code generation. The description (or  
model) has to provide sufficient information to drive the generator.

That said, I'm not at all sure that generating code for RESTful Web  
APIs is a good idea -- but then again, I'm not sure the description  
language itself is a good idea either.

I find it hard, though, to imagine a useful Web description language  
that would not also be usable for code generation.

Stefan

> Marc.
>
> ---
> Marc Hadley <marc.hadley at sun.com>
> Business Alliances, CTO Office, Sun Microsystems.
>
>
>

Received on Friday, 17 June 2005 21:59:00 UTC