Evaluate purchase of Google Workspace for InstructLab.ai domain

[lhawthorn]

We are considering using Google Workspace as part of the collaborative infrastructure for the InstructLab project.

This issue is related to #89

[lhawthorn]

Capabilities required for collaboration tooling which we currently believe will be well met by Google Workspace:

• Calendaring tool (community meetings scheduling and sharing of dial-in / video conferencing support; notification of events where InstructLab will be discussed or where people can meet maintainers or e.g. meetups)
• • Ability to embed this calendar online in GitHub and on other web properties
• Related to above, video conferecing/dial in mechanism (i.e. Google Meet)
• • Video conferencing solution must support the ability to record and then publish meeting recordings publicly (prefer easy integration with YouTube)
• Email lists (Google Groups) as detailed in #89
• • Ability for anyone to subscribe to these email lists whether or not they have an instructlab.io email address
• Google Drive for document and other non-code artifact storage
• Tooling to create non-code artifacts, e.g. presentations, drawings, spreadsheets, etc. ** Ability to share and allow modifications to any of these artifacts whether or not the user has an instructlab.io email address

Assuming Google Workspace can meet all of these criteria, we will also need to choose the pricing plan that supports these options (appears to be Business Standard) and who the administrators will be.

Proceeding under the assumption that we wish to have an administrator group of ~10 people the cost would worst case scenario be 14.40 USD/user/month or 1728 USD per year. We may also be able to negotiate a discount since Red Hat is already a consumer of Google Workspace.

[lhawthorn]

Asking @cybette to vet list of requirements with InstructLab teams and, if we agree these requirements are sound and met by Google Workspace, @lhawthorn will go find budget for this cost.

I believe it may be possible to negotiate a discount to 12 USD / user / month, which would mean cost for 10 admin level users would decrease to annual spend of 1440 USD per year.

[jeremyeder]

Where is the number 10 coming from? Is that the list of admins only? IOW if community folks have a google account of their own, we don't need to pay for them, right?

[lhawthorn]

Admins is everyone who is currently listed as responsible or consulted in our OSPO workstream, namely:

1. yours truly, @lhawthorn
2. @caradelia
3. @cybette
4. @jjasghar
5. @joesepi
6. @mmcelaney
7. @bradtopol
8. @mairin
9. @jeremyeder
10. Susan Malaika (IBM - focusing on events)

While we could likely make this a shorter list and some of the names on it would not be correct for the long-term, I am hoping to avoid a situation in which we are dependent on someone being available at an off hour to just create a silly Google document for people to use to collaborate on or to add a meetup to the calendar or to spin up a Google Meet to have a call with community members should there be a desire to talk real-time.

And to specifically answer your question, @jeremyeder, we envision a world in which anyone who is a user with a Google account can use certain resources (contribute to a Google document, join a Google Meet session, edit a presentation stored in Google drive) but admins are the only ones able to create certain resources and manage permissions to them.

So, J. Random Human cannot create calendar invites on the community calendar, but would be able to add text to a Google document already created by one of our admin users or join a Google Meet session set up by an admin user.

Does this answer your question sufficiently?

[jeremyeder]

"envision a world" -- just trying to confirm that regular exampleuser@gmail.com can participate in this workspace. if thats confirmed, and you have the $2k/yr budget then I say go.

[lhawthorn]

Yes, that's correct, exampleuser@gmail.com should be able to participate, though they may not be able to create various items. And we'd need admins to add them to resources, e.g. request access to this Google doc.

I don't have the budget, but that doesn't mean I won't go find it. :)

[spotzz]

My only concern, which I really don't think we'll run into with the target audience here, are the folks who will only use Open Source. From the capabilities listed I think it might actually save money if we're not hosting and managing ourselves. Not including admin as this still needs admins.

[cybette]

I believe it may be possible to negotiate a discount to 12 USD / user / month, which would mean cost for 10 admin level users would decrease to annual spend of 1440 USD per year.

The 14.40 USD rate is for monthly billing. I'm quite sure we'll want to go with annual billing which brings the rate down to 12 USD / user / month.

[jjasghar]

I mean we could set up a jitsi instance if we are worried about google meet. (I got one working at https://meet.MYLASTNAMElabs.io)

It's not horribly hard to run. mailmain mailing lists, jitsi for meetings, and matrix for chat? 🤷

[jjasghar]

https://matrix.org/blog/2020/04/06/running-your-own-secure-communication-service-with-matrix-and-jitsi/ it works really well for FOSDEM.

[spotzz]

I will say we've been using OpenInfra's jitsi instance for the OpenStack PTG this week and a lot of the issues we had originally seem to be gone. It's been working great with large numbers of people participating, raising hands, etc. I know our configuration doesn't record but I think you can get that functionality

[cybette]

We could piece together the separate tooling for meet/chat/mailing lists, but there's also calendaring and drive/file sharing etc. so ideally we want something that pulls everything together.

Perhaps Nextcloud?

(don't get me wrong, I personally am a huge Jitsi/Matrix supporter and would be happier if everyone is on matrix so I don't have to log in to several slack instances!)

[cybette]

Update: checking with procurement how we would proceed with costs, payments, etc.

[lhawthorn]

Folks, my primary concerns with this decision other than meeting the requirements as outlined above:

• Get it done as quickly as possible, we are facing down an aggressive launch deadline
• • this does not mean making a bad decision in haste, it does mean we cannot admire the problem together for too long
• Choose a suite of services that will be easy to hand over to a software foundation home when InstructLab goes to a foundation home
• • Moving InstructLab to a foundation home is currently part of future community roadmap (which lives in all of our heads and will actually get documented real soon now(TM))
• • Part of the allure of Google Workspace is I know it is a well understood platform for every foundation I have worked with in the past, other sets of tools are not as well understood AFAIK
• Choose a suite of services that are well understood and often used by folks
• • We are appealing to a different set of developers than Red Hat has catered to in the past and, frankly, I am not sure how many of them care if we use free and open source tooling. I care quite a bit, but I also do not want to create friction for people to engage with us by having them learn yet another tool for simple collaboration work. If we do make an explicit and intentional choice that we will use as much FOSS tooling as humanly possible for InstructLab - GitHub is, notably, not FOSS - then we need to have an intentional dicussion about that as part of our project value system and document that we made this choice explicitly and why. I have zero issue asking community folks to do more work to learn new tools because part of our values as a project is an open source LLM ecosystem that is managed with open source tools, but then we need to make that something chosen intentionally and that we are all willing to 'defend' as a choice. (PS - I <3 Nextcloud and Jitsi.)
• • I worry that we are going to look like hopelessly out of touch and insufficiently cool to the average developer working with LLMs or AI anything because we are using Mailman or Jitsi or anything they didn't grow up using. Explicitly stating my concerns because I want this project to succeed, and people get particular about the personal identity related aspects of participating in a project, and which tools the project uses are a part of that.

We need to meet people where they are if we want to win them over. (But did I mention I <3 Nextcloud and Jitsi?)

[lhawthorn]

For something of this low cost, we can swipe a p-card for now and do the correct needful later. I believe OSPO Community Infra team has a p-card.

Outstanding related actions to make choosing a Google Workspace suite of tools possible (if that's our choice ...) @cybette to track down who owns the InstructLab.io domain (think this is IBM), initiate process to transfer ownership to Red Hat (Red Hat OSPO community infrastructure team can help), as I believe that unless we own the domain getting Google Workspace access for it is not within our power.

[jjasghar]

Get it done as quickly as possible, we are facing down an aggressive launch deadline this does not mean making a bad decision in haste, it does mean we cannot admire the problem together for too long

We are appealing to a different set of developers than Red Hat has catered to in the past and, frankly, I am not sure how many of them care if we use free and open source tooling.

I do want to say that for communication (and frankly "trends" in this space) Discord has become a standard channel for this. I do want to say I don't want to use it, but it does seem to be "what the cool kids" are using now-a-days.

I felt if I didn't say this we'd be doing ourselves a disservice in this issue as a whole.

[lhawthorn]

I understand and hear you re: Discord. I have not brought up Discord because folks have seemed pretty pro-Slack in the discussion thus far.

[jasonbrooks]

I understand and hear you re: Discord. I have not brought up Discord because folks have seemed pretty pro-Slack in the discussion thus far.

In addition to the AI-community inclination toward Discord, it's free (you need to pay a nominal amount to give the server a custom name, we went through this w/ lightspeed initially), and it's much simpler to manage the users. Slack needs an invite mechanism, and it can be awkward. For instance, Pytorch uses a google form to try and vet whether users are slack-worthy: https://bit.ly/ptslack

[lhawthorn]

Thanks to @jasonbrooks we have purchased 10 seats at business user level for instructlab.ai, which is the actual domain selected as opposed to .ai as appears in the issue title. Closing this issue.

[lhawthorn]

@instructlab.ai email addresses created for:

Admin @cybette @lhawthorn @joesepi @caradelia

Working on getting the details required for our other 5 seats to share with @mscherer

[lhawthorn]

We have also now created accounts for

@jjasghar @mmcelaney

[lhawthorn]

User accounts created for @kelbrown20 @mairin