saving launch paths in database

[bitnik]

Firstly I should clear that with "launch paths" I mean the data in the optional file or url path field in binder form.

Proposed change

When a launch happens, some launch data (namely image, token, repo_url, binder_ref_url, binder_launch_host, binder_request and binder_persistent_request) are saved in user_options field of Spawner table. And AFAIK for example JupyterHub uses this data (for named servers) to re-launch user server.

image, token and repo_url are also really useful if you have a BinderHub with authentication (also with or without named servers) and want to remember exactly what user launched last time. But launch paths are missing and this means that we cant really start the user server same as user did. For example if user launches with /lab url via binder form and later starts the server via the button in JupyterHub home page, user will be welcomed by the notebook UI, not Juypter Lab. So thats why we thought it would be good to save launch paths together with other data in user_options.

Alternative options

This would be also good to know.

Who would use this feature?

People who wants to deploy a BinderHub with authentication and/or wants to have more/other customizations. For us we would use it in our persistent BinderHub.

(Optional): Suggest a solution

#1114

---

What do you think, about saving them in database and also about how to implement it?

[manics]

urlpath can be used to pass additional parameters to a notebook extension in addition to launching applications, for example:

• https://discourse.jupyter.org/t/tip-embed-custom-github-content-in-a-binder-link-with-nbgitpuller/922
• https://discourse.jupyter.org/t/pass-parameters-to-a-notebook-on-jupyterhub-binderhub/1973

Do you anticipate this usage causing problems?
Is this strictly internal to BinderHub, or would it ever become public? https://mybinder-sre.readthedocs.io/en/latest/analytics/events-archive.html

[bitnik]

Do you anticipate this usage causing problems?

First of all thanks a lot for the links, they were very useful to read. After a quickt thinking, I dont think that this usage would cause problems.

Is this strictly internal to BinderHub, or would it ever become public? https://mybinder-sre.readthedocs.io/en/latest/analytics/events-archive.html

This is very interesting. For example, this would be also useful for analysis such as jupyterhub/team-compass#277, but I am not sure about having this data public regarding user privacy.

[minrk]

They aren't really "launch paths" so much as post-launch redirects. The URL is not relevant to the launch, which is why we don't pass it; instead, it's a location to redirect to after the launch completes. Many links to the same repo/ref with different url/filepaths should result in exactly the same launch but different paths. If we stored state so we could see a browser's running servers, for example, it would make sense for each link to open different URLs in the same running instance rather than spawning a new instance for each.

If we want to set things like the default redirect (as when lab is enabled in a regular jupyterhub deployment), I think that's a different feature that should probably exist at a different level, and not the same as persisting a post-launch redirect as a launch parameter.