Decision log

March 29th, 2020

no role management
“Customer agreements are in place with each Auth0 customer. These agreements cover data transfers to third parties that may occur as part of Auth0’s provision of its services to the customer.”
expensive after 7000 users (and we hope to scale)

When a user clicks on a tag, we should show a sample of resources/hangouts/learning paths/notes/projects/groups associated with that tag (paginated)

In essence, tags are “global” — and should be editable/creatable by a moderator population

We looked more into the docs for django-taggit and researched using that in conjunction with DRF.

ease of use
because a global tag manager makes more sense for our uses cases (global tags that would span across models)

We selected Django with the Django REST Framework because it:

provides a common framework for building REST services
provides an opportunity for those learning Python to contribute
allows those experienced with Django to build on skills they already have
provides OpenAPI schema/documentation of the services
exposing CodeBuddies as REST endpoints encourages decoupling from a single UI
while other technology choices may offer better performance in certain situations, we haven't identified any CB functionality requiring performance optimization

We considered but ruled out the following:

Go - low interest among the CB community to learn it
Java - low interest among the CB community
Node - some interest, but not enough people familiar with it
GraphQL - additional complexity that doesn't solve an identified problem in the CB codebase. Can be added later if required. Isn't performant with Django right now, though.
Serverless and GraphQL (using Netlify and AWS Lambda) -- newer/untested technology, fewer Stackoverflow references, depending on Netlify may be expensive if traffic grows

(h/t to @BillGlover for writing most of the bullet points above)