Initial observations based on team discussion on 20 June 2013
?
Notes from John Lowe
Data model, data content
- In current CineFiles, user can see most values ("related records") in the "main screen", don't need to switch to a tab...some screens have a button to bring up a separate screen for "relations".
- Some readonly fields show data from other tables, e.g. "other directors for a film"
- Entity-Relation diagram
- Make a big one. Glen will send Lam the file to render to PDF. We will post it here.
- Vocabularies are all stored in individual tables with an FK relation to collectionobjects (=documents) ...LCSH sub heading, ISO country names, person names, organization names.
- pfa_denorm2: maybe use this db for migration? (about 500k to 1M rows).
- (John: but is all the data there? It seems a lot of the metadata about the data is not: e.g. dates of update, who did the update?)
- 2 scripts: one looks for records that have changed...one regenerates denorm altogether.
- films table is a controlled vocabulary of film titles.
- reports? several "courtesy reports" exist.
- hierarchical searching? not much!
- images
- 500K images (each page is an image);
- image creation has slowed down...
- pages (images) are stored on the webfarm, organized in the files system by docid. Look ma, no database! just a directory structure.
- relatively static collection, at the moment.
- They may want a DAM...
- documents are the collectionobjects
- is a page a media record? if so, how do we order pages...?
- are films citations?
Migration observations, options, and suggestions (technical)
- Consider some sort of intermediate warehouse for use in the migration (see pfa_denorm2 above)
- Could we simply generate exactly the same 3 denormalized tables from CSpace, and then have the existing portal "just work"? use existing jsp, which is based mainly on views?
- Perhaps we should try this: copy the tables to pfa-dev (when it exists), and get the exising JSP webapp working there.
- Then the task of supporting the public portal comes down to making denormalized versions as we are doing for Delphi and (probably) UCJEPs.
- Customization
- lots of sysbase stored procedures => need for event handlers and batch procedures
- Glen will get stored procedures and triggers and put them someplace for viewing
Customer involvement
- this is really a research database. specialized. used by film researchers, not casual IMDB users!
- who will we be training?
- we dunno! people glen trained are gone; there are a few part-time replacements
- Chris: looks like Nancy and one other person who is helping with the cataloging; no other SMEs, users.
Other speculation
- ResearchHub
- What about a retrospective / lessons learned on the ResearchHub effort?
- Can see an eventual need for "integration" with ReasearchHub.
- They had lots of fields, we mapped them to 8.