RFC-1: Decide on stack

[doridori]

https://stackshare.io/

Initial Ideas

graph DB

• Neo4j
  • GraphineDB
  • https://github.com/neo4j/neo4j-browser
  • GPLv3, if run server side seems harmless (link)
• Amazon Neptune
  • How to guide to access in a serverless way https://acloud.guru/learn/aws-go-serverless-graph-database
• Amazon Appsync https://aws.amazon.com/appsync/
  • GraphQL proxy
• Cayley*
  • Truely open src
• Azure CosmosDB
  • NoSQL but supports gremlin
  • can be deployed serverless https://azure.microsoft.com/en-us/solutions/serverless/#solutions

non graph

• Firebase RealtimeDB or Cloud Firestore
  • not graph but can roll out serverless solution here
  • JSON or Document respectively
• IBM cloudant https://www.ibm.com/cloud/cloudant
  • serverless with auth
  • https://cloud.ibm.com/catalog/services/compose-for-janusgraph
  • beta

Host (PaaS)

For simplicity need a PaaS, dont have to worry about any real admin

• Digital Ocean
• Linode
  • better than digital ocean
• Heroku*
  • Seems best rated and good support
• Google App Engine
• Amazon Elastic Beanstalk

Host (Serverless)

• Amazon
  • https://aws.amazon.com/cloudformation/
  • https://aws.amazon.com/serverless/?nc2=h_ql_sol_use_servc
• IBM
• Azure
  • https://azure.microsoft.com/en-us/solutions/serverless/#solutions

Host (no host!)

Use file syncing like dropbox and or google drive. Web apps would be awkward, electron app for desktop? Would mean no costs at all for me.

• Google Drive https://developers.google.com/drive/api/v3/about-sdk

server side framework

• Springboot
• Kotlin (Ktor)*
  • simple, and its kotlin!

front end (options)

• Vue
• Service workers (if heavy weight web app)
• Snowpack (front-end build tool)
• React.js (or Preact + HTM)
• HTML5 Canvas / http://graphalchemist.github.io/Alchemy/

OR

Multiplatform

• Flutter!
• KMP with Kotlin/JS
• Electron (no mobile?)

Auth

• Firebase Auth
• Amazon Cognito

[doridori]

Playing with the neo4j local server I think it would make sense to not visualise the graph, and rather display just one level deep from the current point in the graph, max of 10-20, ordered by how many children those nodes have. This would be simpler to read from a users point of view, and visualisation would be useful for admin cleanup only.

This means can cut out the visualisation stuff from the above stack.