ALEG

What I've done

• Loaded AMLC agents part 2

• Added function to allow review of changes selectable by paremeters: who (or all), when (last n days), type of change (add, delete, update, all), type of topic (agent or work)

• Lots of changes, bug fixes (especially in the 'first known date' area) and many tweaks of stylesheets and templates

• Agents recorded at the expression level are now searchable at the work level

• Implemented a 8-step "Merge Agents" user interface. This has been built as a general "merge any topics" system, but until it is given a good test in its agent "configuration", I won't implement the extra code needed to merge works, concepts, etc.

• We had several unexplained update failures. In these cases information from the log strongly suggested that an update was being reapplied, normally with some extra updates. The updates failed because the update transaction includes the old value and new value of each updated attribute, and the old value (as supplied by the web client) did not match what was stored on the database. To try to identify the problem at an earlier stage I made small modifications to the client and server code to hand out a unique transaction number when a client takes a topic for update. The server checks that when the update transaction has been received the transaction number hasn't been repeated and matches the transaction number expected.

• Started coding the BAL data loading program, but sunk into despair/discussion due to the complexity of data represented in textual form requiring careful analysis to identify structures and load/merge into the database. After some discussions, Ben and his team are manually starting work on about 20% of the BAL list whilst I labour on with a program which will attempt to load the easily interpreted records. We'll have a better idea of how this will all work by the end of next week.

• Add general edits based on XSLT Schema (the Schematron approach).
• Revisit output formatting (especially to show all the newly added attributes and relationships) and search screens.
• Drama data - work out what we need to store to record a "performance"
• AMLC bib data load

• BAL loading

• Investigate a problem with diacritics (AMLC bio's) being mangled when updated

• Discuss the approach we'll take to the Lu Rees non bibliographic data (probably the week after next when Megan (an EAD expert!) completes her work in Specials

• Enhance expression/manifestation selection search

• Remove the test works and agents added during training

• A great deal of use was made of the data maintenance screens this week. There are a few unexplained problems (see above) which I think I've corrected, but we'll see what happens this week. I don't think any of the remaining problems are serious.

  On the other hand, a renewed and detailed look at the BAL data (which is required when you start to actually write a program to process it!) revealed a great deal of rich information held within the text fields, especially the edition info. This is going to be complex to process accurately and completely, and I suspect we'll end up with a system which loads the obvious and relies on humans to analyse the complex.