Support for accounts

[okennedy]

It's boring, but if people are going to use this we need at least a limited ability to perform data governance tools. Basic HTTP Authentication works, but is decidedly all-or-nothing.

[mrb24]

A few points that came up at SIGMOD about governance:

• [ ] the ability to share projects with other users
• [ ] live colaboration
• [ ] groups & group projects
• [ ] account administration

[okennedy]

@lordpretzel @julianafreire @heikomuller @soniacq

Given the strong call for it at SIGMOD, it would be worth starting to think about how we would go about implementing this. It seems like a pretty substantial change, so I'd like to have a head-to-head at some point to figure out how exactly this is going to be implemented.

[julianafreire]

These are important features. However, IMHO we should first make sure that we have a single-user tool that is easy-to-use, effective for different (and representative) use-cases, and scalable.

[okennedy]

I agree that ease-of-use and effectiveness for specific use cases have to come first. However, I think you're getting at a deeper question: How do we expect individual users to interact with Vizier. Almost since the start of the project we've been using a tiered design (UI / API / Mimir) that is not really ideal for non-technical users to set up and run. In general, that's been ok, because nowadays most people seem to be looking for hosted tools anyway (in particular if their org can self-host), but this is a deeper discussion: Are we building a hosted tool, or a desktop application?

Hosted

• Easier to obtain scalability by plugging into an available spark cluster
• Vizier presently requires 3 separate servers and a 30 minute install process
• Enables useful vizier features like workflow and snapshot sharing
• Easier to get new users started

Desktop

• Safer to use with private data
• Works offline (though note: Several features like the geocoding lens and mapping still require internet access)

[okennedy]

Moving this issue out of Vizier core.