12:04:53	 From Michael Hildreth : Zoom link for joint session with CompF5: https://olemiss.zoom.us/j/91422766990?pwd=bGJyTHM2NnllVDcyTkpMajR5dDZ0dz09
12:17:53	 From Benjamin Weaver : Question: I usually think about Continuous Integration in terms of running unit tests. How does that apply to the *preservation* of code?
12:19:26	 From Michael Hildreth : It ensures that your code is in a repository, at the very least.
12:20:08	 From Kyle S Cranmer : Yes, we are kind of repurposing that infrastructure. Normally when you commit code and CI runs it needs to build a docker container, install code, etc. for running unit tests. We basically then save the container
12:20:28	 From Michael Hildreth : Also, if you run more like a full pathway rather than a unit test, you’ve then preserved your analysis chain
12:21:33	 From Kyle S Cranmer : +1
12:23:26	 From Benjamin Weaver : I’m missing something then: how do you validate the output of CI? It’s an *analysis*: the result is not predetermined, unlike a unit test.
12:26:06	 From Tibor Simko : CI workflow was used mainly to produce analysis docker image, so validation can be simply to check whether container image contain everything... Also the workflow can run on a small data file to make sure thinkgs a are running fine
12:28:40	 From Benjamin Weaver : But what is the advantage to *not* doing that on a laptop?
12:29:24	 From Kyle S Cranmer : The main thing here is convenience and building infrastructure to preserve all the analyses in a systematic way
12:29:44	 From Michael Hildreth : you could.  The procedure is more important that the platform
12:29:51	 From Kyle S Cranmer : We want to capture the analysis workflow that may be developed by a large group of people.
12:30:29	 From Benjamin Weaver : OK, that is making more sense now. Thank you.
12:30:58	 From Kyle S Cranmer : So when a member of the team commits some code the infrastructure builds a docker image that captures all the environment etc. needed for that step of the analysis workflow
12:32:03	 From Benjamin Weaver : Indeed, so in these tutorials, are there teams, or are all participants a team of one?
12:33:13	 From Kyle S Cranmer : There are example workflows that have many steps, and participants learn how to interact with the infrastructure… change a component, rerun the entire workflow, etc.
12:34:25	 From Benjamin Weaver : I’m trying to say that an important lesson is all the different ways that other team members can break your code.
12:34:48	 From Leonora Vesterbacka : The idea is that participants can learn and bring back the knowledge to their analysis teams
12:40:26	 From Matias Carrasco Kind : Q for Leonora and others, you mentioned 5 years old data, where you draw the line between usefulness/work needed to go even older? Is 10 year old data? 
13:23:39	 From Kyle S Cranmer : https://soundcloud.com/kyle-cranmer-73980343/steinberger-comments-on-open-data-2009-11-03