ALEG

What I've done

• Dataloading

  1. SA Women Bib - completed alt and pseud loading by hand, loaded gender, birth/death information. No bib information has loaded yet.

     Nothing more has been done with the manual processing of the subjects for the Australian Responses to Asia bib.

• Miscellaneous queries and adhoc data analysis reports, mainly associated with 11,000+ works without worktype/form/genre classification.

• Brief meeting with Creagh Cole and Ross Coleman from SETIS about linking to full text of secondary works.

• Added affiliation and affiliation note as agent attributes - maybe this will help in the building of some views such as SA Women where the other agent data does not clearly identify them as belonging to a view.

• System

  We finally cut-over on Thursday evening to the new hardware. We finally managed to ugrade Oracle to version 8.1.7. The new system is running the latest production versions of all the main software we are using:

  • Xalan J 2.1.0
  • Xerces 1.4
  • Apache 1.3.20
  • Tomcat 3.2.1
  • Java 1.3.1
  • Oracle 8.1.7

  There were a few minor problems Friday:

  1. The IE5.5 browser was caching data from the server marked as not to be cached/expired! On investigating, Tomcat 3.2.1 was sending the correct headers, but in a slightly different order from the previous version (Tomcat 3.1), and from Apache. Apparently, this is enough to make IE5.5 dishonour the no-cache directives!! Apache was sending them in an "acceptable" order, so I reconfigured the web system so that all requests now pass through Apache to Tomcat rather than going direct to Tomcat.

  2. Some of the paremeter passing details to Xalan 2.1 stylesheets are slightly different from Xalan 1, which caused merging of expressions and manifestations to not work until it was fixed.

  Load is relatively light on Fridays, so Monday will be a better test, but the new system is operating OK. I will tweak some Oracle parameters for next week to improve memory utilisation - we want to use all the memory, but not start paging or swapping the programs, Java heap or Oracle buffers.

  The system has just over half the memory of the loan system, but it should be adequate if we use it carefully. Dividing up the available memory between the Oracle database, the Java application and its memory cache and the operating system file buffers is an area which needs attention.

  Our new system has a single very fast processor, compared to the loan system which had 6 processors of about one-third the speed. The new system also has faster disk and memory access. This should mean that most requests will complete faster, but if there is a lot of concurrent accesses (eg, many people concurrently issuing complicated and long requests), the new system could be slower than the loan system. The goal is to process each request as fast as possible to help reduce possible contention.

  We have 2 unresolved hardware/operating system issues with Sun:

  1. The SparcIII chip in these early Blade systems has a hardware bug. A workaround patch has been applied, but it reduces performance by an estimated 5%+ (stories here and here and here) As soon as the bug is fixed, we want Sun to replace our processor chip!

  2. After the second disk was installed and the important parts of the system mirrored on the disk (so as to survive a single-disk failure), the system would not recognize all of the disk partitions on reboot. Sun have supplied a workaround which disables hardware assist for a particular and common type of memory operation (set use_hw_bcopy=0 in the /etc/system file). No patch is yet available, but obviously we are keen to get one so as to utilize the full capabilities of the hardware we have purchased.

  The documentation server contents have been moved to the new system. The address of the documentation server is temporarily: http://austlit.lib.adfa.edu.au:7777/. When a new series of domain names for AUSTLIT are acquired, this site address will move again!

  Fran Cassidy deserves special thanks for greatly assisting the migration to the new hardware - it was not an easy task as we took the opportunity to upgrade all the software as well, and without her expert help it would not have happened as smoothly as it did.

• To Do list. Work through the (mostly) minor issues on the to-do list provided by Annette/Kerry/Marie-Louise.

• Finalise SA Women Writers Bib load - the actual bib data!

• More time on planning session on the new user interface with Marie-Louise and Annette.

• Some reports for Kerry and Annette on works without meaningfull type/form/genre classification - these came from BAL/AMLC/WABIB with designations such as "fiction" or "sundry", and we need to work out how best to tackle their classification.

• First known dates.

• The new thesaurus is expected to be available for loading next week - I'm going to have to revise the hierarchy structures to allow topic searches to include words from across the thesaurus "tree".

• Pick up the ball I dropped regarding the Z39.50 interface to NLA holdings.