[Gusdev-gusdev] GUS WDK architecture meeting summary

[Steve F. <sfi...@pc...>]

Folks-

We had an informal first technical meeting of possible core developers 
of the next generation GUS WDK.  The goal was to set some architectural 
foundations and get oriented.  We covered the part of the WDK that 
specifies web sites.  We did not cover web services, which is likely to 
be largely independent of this part of the WDK.

As this was a technical meeting, we did not cover scheduling or human 
resources issues.

Included in the meeting were
  Adrian
  Arnaud
  Angel
  Dave
  Jonathan S.
  Steve

We converged on a number of ideas:
  1. use struts/jsp
  2. continue using a configurable JavaBeans based model
  3. the model will be similar to the current WDK's model.
  4. however, its configuration file(s) will not include any view 
configuration, as does the current WDK config file
  5. the view will rely heavily on custom tags to provide renderers for 
the data types found in the model
  6. we need to figure out how to aggregate custom tags to provide 
 higher level re-usable view elements.  we might use tiles for this.
  7. we will look into JavaServlet Faces to see how it can help us 
organize our view design.
  8. at the outset, we will author JSP pages to provide the framework 
for the view.  in particular, we expect to start with dedicated pages 
for top-level records such as Gene, RNA, giving these pages a hand 
edited look.  These will call out to the tags or tag aggregates.   We 
will ultimately evaluate whether that kind of hand authoring is becoming 
unscalable.  If so, we will add a facility to generate those kind of 
pages from the model, allowing for minimal configuration as needed.  We 
call this facility the rendering engine.
  9. As with the current WDK, the model includes the queries that are 
available to the users. [I am proposing that this part of the model 
could be called the Questions model.  The other part of the model has 
Answers, such as Records and ResultSets.  The model will also include 
ComplexQuestions.  I am steering away from the word "query" because it 
is over-used in this context (Answers use queries to format themselves).]
 10. a good way to get rolling is to produce a UML spec. for the 
Questions model, and then to make a view.
 11. we expect to rely on struts for stuff like logging, error handling, 
connection pooling and possible process pooling
 12. we expect to continue to use our own sophisticated facility for 
Question parameter validation (a la CSP) rather than struts's validation

The action items were:
  1. angel: give a report on JavaServlet Faces
  2. adrian: give a report on the usefulness of tiles given our expected 
use of custom tags
  3. adrian: put together a UML spec. for the Questions model

Steve