CorTeX 1.0

[dginev]

Currently, cortex has no authentication for its workers. Any message that has a correct zmq envelope signed for service "X" will be accepted as a valid request/response for that service.

That is fine for the single-deployment situation we've had so far, but is unrealistic in practice. I would like to extend with:

• An external developer uses the Cortex UI to register via an external account management service -- since I don't want to implement account management in cortex. (Decided on Google OAuth as the biggest / easiest to setup provider).
• The cortex admin should have a minimal dashboard matrix where each GitHub-registered name has a checkbox against each corpus, and also a separate matrix for each service. The cortex admin can then grant corpus rights and service rights to individual accounts.
• OAuth2-based workers
  - a registered external developer uses the web UI to receive a secret, added to the configuration of the workers.
  - The workers will then request session tokens based on the secret from the server, and pass them on in the zmq interchange, to prove their legitimacy to the dispatcher
  - The dispatcher should be able to reject invalid requests/responses silently and instantly, to be robust against various (largely accidental) attacks, where the protocol gets flooded with noise.
• Linux service that can auto-start the cortex frontend and dispatcher on system boot, or at least having a single command (currently dispatcher and frontend get started separately, but in most demo deployments they will be on the same machine and can have a simple unified bootup).
• Preview captchas should be configurable on/off, since only arXiv requires this level of strictness. Also, the captcha guard needs to be configurable, as currently it is linked to my personal token.
• Registration, extension, removal, of corpora needs to be possible through the web UI
• Creation of services, as well as registering/removing them from specific corpora needs to be possible through the UI.
• Clear documentation for first-time installation and configuration, with a minimal number of steps - at the very least the arXiv-to-HTML setup needs to be provided as an example set of steps from downloading cortex to finishing a full latexml run.

This about covers my expectations for a 1.0 release. Wanted to record so that there is a clear path forward - and if anyone is interested in using cortex for their own projects, feel welcome to comment and contribute!