Brainingstorming Notes

Rob's group

It was noted by the group: Workflows differ from services in that they can be changed, but the distinction is blurred

It was pointed out by Alan: Not all workflows and services are determinable

Overall consensus is that portal should be able to execute a workflow

Group vote for Instant Messaging resulted in 0 votes

Workflows are not built collaboratively

Alex' group

Consider when to use portal vs. application. Run locally when needed (like visualisation), run on grid/server when possible.

Error messages -> solutions. Give help, not stacktraces.

Rating, trust network. "I'm a five-starrer on myExperiment" should be a good thing to mention in a job interview. Would mean changing current science culture! "Where's your papers?". Rating on "How many are collaborating with you?" or vice versa. Open peer reviewing, bypass journals. Review the reviewers. Reward good citation practice.

Economy system, like Slashdot, earn "points" by contributing.

Google everything. This doesn't mean that we should depend on Google, but that there should be an easy text search that searches everything in the portal. Avoid massive categories and ontologies, just use tags and search.

Annotate and link anything in the portal, including links.

The me-Science. Scientists are selfish, they don't naturally want to collaborate. Two scientists at the same institutions might never talk if not for some 3rd party bringing them together.

Negative result space, this is what I couldn't get to work.

Domain spaces, I'm an expert in biology/astronomy/X, just show me relevant stuff.

Tags, tag clustering, text mining. Use auto-complete to help people choose "right" tags, could include (but not require) ontology terms. New and popular tags suggests changes to ontology.

myGene, club for people interested in a particular gene. There are scientists that are so focused that when a new service comes up, they immediately check if say their favourite gene is there. If not, they'll leave and never come back.

Running Taverna in web browser (AJAX), build workflows online.

Guidelines and tutorials, demo movies like Ruby-on-Rails. Best practices. Gold standard workflows.

Workflow logs, lab logs (stop calling it provenance)

There's always issues with security in todays technology, "No, you can't do that, you need a certificate X and permission Y. What about just do it non-secure and say that it is non-secure? Non-secure space, non-secure grid, etc, should just work.

Easy grid submission, "Add your local grid".

Should run in "every" browser, like IE 5. No Word documents, no PDF, open standards.

No manuals should be needed. "Simple, clean interface".

Software development through mash-ups, "empty" workflows that can later be realized with real services.

Offline usage. What happens when my laptop can't reach myExperiment?

Automatic citation generation, the "Generate Nature-paper" button would find all the references for workflows and data you have used.

Publishing a paper with corresponding published workflow and data on myExperiment, data/wf-sharing could be a requirement set out by funders.

Fear: "People will laugh of me." Choose carefully what to make public, hide the crap, "edit" provenance before publishing (remove bad runs). After sharing with "friend", remove such access as the friend turned out to not be your friend anymore.

Rumours and gossip space.

Find scientists in your field or interest. For instance, PhD^? students from different universities doing the same thing could connect up. Finding project partners. Finding funding. The real boom of the internet was people connecting to each other. Job search, CV/Job space. Virtual professor, both for consulting and.. eh.. funding.

Knowing the concequences in advance, avoid submitting a workflow that will hammer down all services.

Data management and visualisation.

Different myExperiment installations in different labs should be cross-linkable. People will probably not submit all their data to some 3rd party site, but if they can install and run it locally they would start straight away. In fields like bioinformatics and biology, the workflows are not that important and easily sharable, while the data are precious and best kept in a safe.

Ship data to curator. Curator sympathy, they also need help!

Pimp my workflow! Help me improve this workflow.

What about signal/noise ratio? Computer scientists are well known to produce massive amounts of text. Need policing, content vs. comments vs. rating. Some comments should be better than none.

How to get people to buy in? Need incentives. "I publish my workflow. Why? I think it's nice to see it reused". Citing.

Science is about re-use. Why aren't we sharing?