RE: Architectural Ideas

Hi Sean,

>I virtually completely agree with what you describe here, if I
>understand it correctly.

>Jo's idea was to make every "real" mobileOK Basic test depend
>completely on the preprocessor results document. While it adds a
>little overhead it is a tidy design.

Yes, now we catch the idea. We have thought in the idea expressed in
the diagram. 


>But yes the preprocessor has to do a lot of work to prepare for this.
>It has to run tests which are not strictly mobileOK Basic tests, like,
>parsing and verifying a CSS stylesheet. Is this what you mean by
>GrammarTest and non-declarative tests?

Yes, more or less. We have thought that some tests (named as non
declarative) can be executed in the pre-processor step and generate the
test outcome (info to add to the intermediate doc, e.g grammar errors).
In a following step the "declarative tests" would be executed through
the intermediate document via XPTAH expressions.
 


>Yes, I think all of this logic goes into the Preprocessor and related
>classes like the HTTPXXXXResource classes. My intention was to only
>subclass TestImplementation to implement the "real" mobileOK Basic
>tests. That's why the one method that they all implement takes in the
>PreprocessorResults object.


>PS you are going to commit some code changes? I'm personally fine with
>that as I believe we're very much in agreement and sometimes its
>easiest to start looking at each others' changes to sync up on the
>remaining details.

Yes, we would be very interested in it. (But we haven't asked for 
the developer account yet. Is Dom the person who we should talk to?). 
Before we begin to work in this code, we suggest having a stable version
of the intermediate document in order to define the necessary java code.
We can talk this in the next audio.

Received on Friday, 20 April 2007 12:27:00 UTC