2010-06-17 Thursday Meeting
Attending:
Agenda
Accepting minutes from last meeting (no Tuesday meeting this week)
Matters Arising (<10 mins)
-
Andrew: Add SPIM to experiment type!
- Should this go in AcquisitionMode on Channel or Type on Experiment?
- Chris: Donald can't see Squig. Please fix!
-
Andrew: Add SPIM to experiment type!
Review of Paris User's Meeting
Email and Forum Review - checked at (5 mins)
link to last 7 days open tickets
ID Status User/email Type Created Selected file Comment Error
2486 new filipevb@gma... Insight 4.1.1 2010 06 16 - - java.lang.Exception: Abno...
2485 new a.j.d.melvin... Insight 4.1.1 2010 06 16 - - java.lang.Exception: Abno...
2484 new N.Hegarat@su... Insight 4.1.1 2010 06 13 - I was deleting a folder java.lang.Exception: org....
2478 new unknown Insight 4.1.1 2010 06 09 - reproducible now. tried to log... java.lang.Exception: Abno...
2477 new unknown Insight 4.1.1 2010 06 09 - had just entered username and ... java.lang.Exception: Abno...
2476 new unknown Insight 4.1.1 2010 06 09 - trying to connect to a remote ... java.lang.Exception: Abno...
- AOCB (<5 mins)
Notes
-
Matters arising
-
CellProfiler: Licensing & distribution of bio-formats
- SPIM: Goes in AcquisitionMode
- Squig is fixed.
-
CellProfiler: Licensing & distribution of bio-formats
-
Review of Paris Users' Meeting
- Public notes should go up on website
- Any confidential notes should be put on the internal page
- Anything that's small might go in 4.2, all other tickets are unscheduled
-
General feeling
-
Model actions/timescale:
- Andrew: seemed to be mixed messages; "too unstable...but we want X"
-
Chris: Too much responsibility on the implementors.
- Need better help (reference implementation?)
- Andrew: using codegen for other languages
- Jean-Marie: we are quiet for too long
-
Jason: the world needs to know our priorities, but the output of a release is more than the model,
- includes software, validators, libraries, examples, ...
- Jean-Marie: why did some companies go to another university for implementation?
- Chris: did the university not contact them?
- Jason: is a summary, "more updates on what we're intending to do and more information for implementators"?
- Andrew: licensing for reading/writing our file format cheap? Chris: BSD? Tabled.
- Jason: But work we should do? Chris: yes. Like stylesheets.
- Currently, very few sites will use bf to write
- Andrew: had misunderstood the conversation about dc:creator
- Jean-Marie: that's another issue - still need to have a methodology for release
-
Josh: 2 things we've discussed before...
- Having StructuredAnnotations registered as a namespace so that we can upgrade them
-
intermediate, unstable releases that implementors don't upgrade to
- A stable once per year, unstable each month
- Andrew: issue of not being able to open files created with the intermediate releases
- Jean-Marie: need to find a middle ground
- Andrew: bioformats would need to read the unstable versions
- Andrew: we will have to provide an upgrade (Josh: and/or downgrade) path for any intermediate releases
- Jean-Marie: is that too ambitious with our personnel?
- Jason: a month sounds too frequent. 2 months sounds realistic.
- Jason: We could make a plan, the 6 2-month releases between stale releases
- Jean-Marie/Jason: we need the data in the system. have real use so we can test.
- Chris: need to have - codegen, objects, db tables, ...
- Chris: do we target the tiff library?
- Donald: what objects are we targeting? if the library changes, then it would break implementors
- Tabling for next releases. Roughly: Stable release for end of 2010, with new methodologies, fixes/cleanups, etc.
- Will wants to know if the community wants us to spend time on that effort.
- Chris: the commercial community certainly does
- Jason: cF. the "Metadata Matters" paper. We want them writing compliant TIFF. Want to avoid invalid TIFF.
- What's to prevent somone just using bf? Model support, speed/Java, and licensing.
- Chris: might not be in the next release
- Andrew: sooner rather than later
- Could discuss with micromanager for re-use (further decouple on the Java side)
- HDF
- Jean-Marie: is that something we want to pursue? Jason: ok, why?
- Donald: NDIM
-
Josh: things to keep in mind...
- need to balance this against efforts on FS since they are orthogonal
- will need to use JNI library to do things in our stack
- if we have the JNI library in our stack, we need to do installers
- Chris: benefits include whole plate in a file, ...
- Donald: for something like FLIM, you don't want all channels (seek problem)
- Open questions: do we store XML? (Josh: No.) Would we then move from XSD? Etc.
- Jean-Marie: putting it on the table just to be ahead of the pack
- Chris: concretely investigate how much effort its going to be
- If there's a ref. impl. (I/O provided), then perhaps no one cares about the format
- Andrew: conversation with Robert (PE), previously worked on VLDB (features/characterizations)
- Chris: what did they tell us about FS and result storage?
- Will: need more examples of what we have
- Jean-Marie: we may need to get pytables hooked up and break it before we re-implement
- Will: perhaps get a heatmap working
- Jason: eye-candy stuff to exercise it - add heatmap to browser in insight/web
- Donald:
- Paul Barber & FLIM were a highlight. But a script repository is a requirement.
- He also wouldn't be doing rlong.getValue() in matlab. API issues.
- pytables could be excercised by CellProfiler
- Jason
- Credit to everyone for having work recognized by the community
- Nice to have people like Bob Murphy saying this is what we want
-
Concerned that we've worked so hard on 4.2, have so much to do but we tend to be focused on big infrastructure
- Martin (PE): "You need some simple stuff that looks great" (e.g. heatmaps)
-
Europe/US: "I want my group to be subsumed by what you've done but it's complex"
- We could constantly go on road trips
- Our tendency would be to work like dogs until ASCB, but do we need to pick our heads up
- Adamant about having big images solved in some form by the end of the summer (e.g. SVS works in bf, chokes in importer, etc.)
- Planning for road trips around September
- Validation of our work is Gianluigi "it's not about images" (hadoop, xen, model generation, etc.)
- Chris/Josh/Jean-Marie: clean up dead bodies, do some easy/flashy stuff, and then work on penetration
- Jean-Marie: maybe we don't decline too many invitations.
- Jason: something like 2-3 weeks at a couple of places. Chris: probably a good time.
- Jean-Marie: Internally, going through future tickets for easy/flashy bits.
- Jason: an iteration on low hanging fruit, then several on the big stuff, maybe some on the road
-
Things to think about
- Invitation to Bob Murphy's lab, in combination with bio-imaging bioinformatics (Janelia, this year in Pittsburgh)
- also to M. on Atlantic French coast for another meeting.
- Josh
- PSLID/modelling
- Knime/flashy/analysis
-
Work for other clients/fiji. Andrew: Should there be a webpage with 3rd party extensions? Could be combined with success stories.
- Planning: when/how? Jason: Make list. Jean-Marie: But how do we have a healthy, open discussion for priorities?
- FS
- Two camps: people who want to use OMERO as backup and those who want full flexibility with their own data store
-
There are still even folks who want to just dump TIFFs to disk
- Jason: If people want to use native TIFF to use other applications, why don't those talk to OMERO??
- Chris: people who say, "unless if you read my files, I won't use it"
- Doing the simple case. No automated reconciliation.
- Need the manual reconnection functionality, but it's not platform specific, etc.
- Jason: As a user, if I can see the data in the repo, then I should be able to do everything.
- Chris: the only thing you can't do is view it.
- MAJOR issue: permissions/security/user-mapping
- For now, won't be running as root
- And files may get hidden you depending on who imports it
-
Model actions/timescale:
-
4.2 Release
-
Open
- 2010-06 schema added. Need stylesheets. Then test_images_good run through.
- Export. Shouldn't be very complicated.
- Bug from Demo servers
- Config seems solid. Need
- Josh puts in quick fixes for mage.
- Callbacks: Ice bug which prevents Usermode processor and all other callbacks. Might not be too important.
- Database ugprade: on the order of a week. 2 weeks on nightshade after done
- Script workflow: performance of scripts
- deletePlate OOM; 12-13 seconds regardless of size of images. 120 hours to delete plate. Some iterations that won't fit into 4.2.0
- Hibernate performance. (Another point release)
- Chris: need almost a month
- Jean-Marie: people coming in and out. If it takes longer, than it'll be September
-
Donald: still delivering new stuff (matlab/cellprofiler/flim)
- Don't have matlab version for the latest CellProfiler 1.x version
- Michael evaluates Matlab bindings. QA build tomorrow for him.
- Two reports about Windows not installing on 4.1.x. Josh needs to track down the leads.
- RC end of next week (June 24). and upgrade nightshade for
-
Open
Actions
- Andrew starts a TODO list/schedule (jul1-dec15) of methodologies (plone) and pass off to Jason