Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Has been sitting in on discussions with Sun to set up test harness.  Current plan is to have 3 machines.  One running Fedora, one running MySQL, one running the test suite.  Goal is to get higher performance.  Try to determine where there are benefits to having certain types of hardware.  Also seeing what can change hardware-wise to optimize things.  Sun wants to know what the recommendations are for people using Fedora on Sun hardware.  32GB ram, 10-100TB disk.  Will have the lab accessible to Fedora Commons.

Also fixed fedora-find bug.

Kai: interesting.  So far, our tests have been on commodity hardware.  Glad to help with input.  Wonder if it would be useful to create benchmarks.  If there's one set of data we can use across environments.

Chris -

On Friday, gave talk on Fedora basics and upcoming
features/interesting stuff.  Pointed to wiki (now publicly visible) to orient new
developers/committers: http://fedora-commons.org/go/fcrepo  With the
new website, this is where the link for "Fedora Developers" or similar
will go.
Asger Blekinge-Rasmussen (State and University Library of Denmark) gave a
presentation on his Enhanced Content Model work:
http://fedora-commons.org/confluence/x/SICn and it looks very
promising.  They have also developed an interesting web-based object
builder tool (they contracted out to do the work, but will be making
it available via Apache2).  Want to help as much as possible on
getting the community's eyes on this and getting their feedback.  We
discussed the possibility of bundling this somehow, or making a link
to it prominent on a future Fedora release as a means to get more eyes
on it.  It probably won't be ready for any kind of bundling in time
for Fedora 3.2 (early spring), but likely sometime after OR'09.  Still
more discussions to come on this.  To aid in his development, Asger
has also written an improved set of documents on API-A and API-M and
is now helping to port it over to our site so others can benefit.
Matt Zumwalt gave a demo of MediaShelf's ActiveFedora.  This is
similar to Ruby's ActiveRecord concept, and the basic idea is that it
makes working with Fedora objects very easy for Ruby developers.  The
model of what you're working on is specified in ruby code, and the
idea is that it could be (but isn't currently) persisted in a Fedora
content model object as a description.  Matt has also had some
discussion with Ben O'Steen who is doing a similar thing, but with
Python (possibly using JSon as the format in which to describe the
model?).  They had been thinking that it may be possible to converge
on a single format (rather than making it Ruby-only or Python-only)
and will be investigating this more.  There was also some good
discussion on the differences between the ActiveFedora approach and
Asger's work.  It was noted that they came at it from very different
angles; Asger thinking more about adaptability and use in many
contexts, and MediaShelf thinking more about ease-of-use when putting
together UIs quickly.  Matt is also hoping to have a "Fedora-backed
Application in 15 minutes" demo for OR'09.

...

Eddie - word+sword/dev8d update, mulgara lib update

Just talked to Glen Robson.  Did orig SWORD impl for Fedora 2.2.  Looking to finish the work to a) upgrade to SWORD 1.3 and work with Fedora 3.1 by end of month.  Currently nothing in their subversion on that project.  Sword folks want to move toward using AtomPub.

Mulgara 2.0.8 update, looking fine before conf.  On hold, what to do about native store.  Use new XA1.1 as default?  Not as well tested, will force a rebuild.  For now, maybe makes sense not to worry about it.  Can go either way at last minute if we need.  Will probably be another point update of Mulgara.  Choice btn old and new storage is a config in the jar.

Gert - likes the public committer notes, mostly listening today
Kai -main interest is in performance.

...

Kai: Fedora triggers garbage collection after each ingest.  Has done large ingests and monitored heap.  Gets much better without explicit triggering.  The 1.5 JVM really does a much better job at GC.  No problems.  Action: Add to tracker.

http://fedora.fiz-karlsruhe.de/docs/Wiki.jsp?page=Java%20Heap%20%26%20GC%20TuningImage Removed

http://fedora.fiz-karlsruhe.de/docs/Wiki.jsp?page=SqlImage Removed