Why branches?

[matthewjosephtaylor]

Unless I'm mistaken git data store internally can store any random bit of data.
Having a branch accessible via normal git feels like a rake as users inevitably update the 'events' (which I'm imagining will be presented as files?) inside the branch.

Why not go bigger and create your own type via another kind of ref? Not a git expert but feels like that should be possible, perhaps even backwards-compat with other git-impls just ignoring the 'junk' in the data store?

Love the idea of a decentralized issue-track/coordination/discussion meta layer ontop of git. Also wiki would be excellent (though the github markdown-file branch maybe works for that good enough)

Do think the concept of a standard way of handling this inside git vs an outside system is a worthy goal.

Good luck to you on your adventure sir! Let the haters hate. The journey is the destination. :)

[sineptic]

I guess he want to be able to sync all things with plain git and implement comments/issues etc. on top of it, not creating his own everything. It would come with a bonus of being able to use jj.

[xeubie]

I updated the idea here: https://www.youtube.com/watch?v=iBDUTaKWVdU

It would still be a special branch, but it would store the data in the git commit message rather than separate files. The point of being a branch is that I want to reuse the machinery of git commits to create a DAG of events for the server to process. Also, being an explicitly named branch will make it easier for people to push it to different Haxy instances.

[matthewjosephtaylor]

+1 on the commit idea. Elegant solution.

I suppose organizing via branching also allows for the possibility of a per-code-branch tracking. Many ways of skinning the cat of how to do code-branching flows. The beauty of standards is there are so many to choose from :)

In other words the 'haxy' branch that branched from a specific 'code' branch sort of automagically signals it is related to that portion of the codebase. Just a musing but perhaps a kind of interesting potential direction you're contemplating already? Not a single haxy branch but multiple I think does lead into some interesting directions.

Liking where this is heading/seeing your thought process. Excited to see what comes next.

[zkfriendly]

it would store the data in the git commit message rather than separate files.

I mean, it's beautiful!

[strange]

Having it as a branch instead of a magic directory also ensures that there is a single source of truth (instead one truth per branch).