[GUSDEV] Workshop Review & Notes

[Michael S. <msa...@pc...>]

-- GUS Workshop Discussion Review

* Documentation and Usability:
Several additional types of documentation were identified as a means to
making the schema more accessible as well as moving towards a common
semantic understanding.  These types of documentation include an "SQL
Cookbook" for common queries, ER diagrams, and improvements to the Schema
Browser that would allow for web-based updating and modification of table
and attribute level documentation, as well as use cases, in a wiki-type
style. The existing wiki was mentioned as a very good location for user
submitted comments, notes, etc. It is anticipated that documentation would
flow from the schema browser, wiki, and other sources to the official User's
Guide, making that resource increasingly valuable as a single point of
reference for using GUS.

Action Item:  Mike will be improving the schema browser to support working
with the documentation.

Action Item:  The hand edited object documentation needs to be improved.

Action Item:  Mike will move the Plugin API documentation (Brian Brunk's
document) into the Developer's Guide.

* 3.5 Migration
Three major approaches to upgrade to 3.5 were discussed: an in place upgrade
using SQL scripts, a migrate and transform process from a 3.0 database to a
new 3.5 database, and starting over with a new 3.5 database and repopulating
using source data.  Almost all groups identified the last option as
unfeasible.  The group agreed  that the ultimate decision between the first
two approaches will depend on the magnitude and type of changes, and that
there was not yet a good grasp on those.  Further, some groups may choose
one approach while others choose another.

Action Item:  Mike will review existing tools that compare schemas and
provide reports.  (Done:  I looked at TOAD and Oracle's OEM, and ultimately
decided that the functionality GUS can provide should be sufficient and
we'll additionally benefit being able to address GUS specifics).

Action Item:  Mike will publish the list of 3.0->3.5 changes.

* Project Management
The group discussed and agreed on the importance of using the "Bugzilla"
issue tracking system for managing bugs and GUS changes.  The development
flow then starts with an issue being created in the issue tracker.  This
issue may have been preceded (in the mailing list) or be proceeded (through
the issue tracker comments) by a discussion on the merits of the change.
The change will ultimately be accepted or rejected by a single individual
who "owns" the component that the issue affects.  (This ownership will be
assigned automatically when the issue is created in the tracker).  Once
approved or rejected, the appropriate changes will be made in the source
repository (applying the patch or confirming the patch for approved changes,
or doing nothing or removing the patch for rejected changes that were
committed to svn).

GUS releases will occur on a more frequent basis, perhaps as regularly as
monthly, with clear upgrade instructions for groups that prefer to upgrade
less frequently.  

The group discussed and agreed to a proposal to manage the GUS Schema and
application framework as separate projects (i.e. with separate version
numbers) to simplify dependencies and upgrading.

Action Item:  Move the GUS schema to a separate project.

Action Item:  Done: A list of schema and component owners have been compiled
and will be sent to the list.

Action Item:  Mike will modify the tracker to support emailing reminders and
email issue submissions.

* Schema Discussions
The importance and wisdom of the placement of the housekeeping columns was
discussed, which possible solutions being changing the requirement (really,
auditing code to ensure that the order isn't assumed), or using a view
layer.

The use of global identifiers was debated, without a clear resolution, but
with a general consensus that more discussion was needed.

A variety of schema clarifications were discussed primarily focusing on DoTS
and SRes. These can be used to frame use cases for documentation purposes.

* Resources Repository
The resources repository was discussed, particularity whether it should be
shared among the BRC groups and/or with the GUS community.

* Tools and Application adapters
Several tools and applications were identified that the community would like
to use with GUS:  Microarray Tools, Manatee, Apollo, Artemis, manual
curation tools, tools for working with clinical data, and proteomic data
tools (Wastling).

* Hibernate

A small working group had a side discussion to get Hibernate running with
GUS.  The first milestone will be generation of mapping files using the xml
schema.  Once complete, a proof-of-concept application will be developed,
which should also help to identify what GUS functionality hibernate will
need to provide.