Obligatory reposting of the same blog article I wrote in 2022: "Cross-fork object sharing in git is not a bug."

https://people.kernel.org/monsieuricon/cross-fork-object-sharing-in-git-is-not-a-bug

Working as a baseball umpire is just like playing video games -- any money you earn you spend on getting yourself higher grade armour.

He said only 100M

RADICLE

Some months ago I learned about Radicle, a truly distributed git forge based on a custom gossip protocol similar to SSB. This allows collaborative code development without the use of any centralised nodes altogether, much less ugly monsters like Github.
See https://radicle.xyz for more details about the implementation.

My experience with it

TL;DR it's almost good, but not quite there yet.

Longer version.

The Good:
Initial setup is easy. Generate keys, run a node, seed your repos, clone others. Despite being fully distributed, Radicle still has a notion of repo ownership, implemented via cryptography. Every repo has one or more delegates, whose versions are considered master copies in case of conflicts.

Unlike other git forges, everything about the repo is the part of the repo. Ownership information, access permissions, PRs, issues, everything is implemented via git objects. You won't ever need to open a browser to submit a PR. Furthermore, you can do all of this while being completely offline. Your work will automagically synchronise once you get internet connection.

For better availability, Radicle has the concept of Seed Nodes. These are (almost) always online nodes with public IPs that donate their disk space and bandwidth for spread others' repos.

The bad:
Bugs. Bunch of them. This is what you get for using software with versions like 1.0.0-rc14. Sometimes my two nodes fail to connect, citing some cryptic error as a reason. My seed node froze up a few times, no idea why.

Radicle is implemented in Rust, which sometimes adds to it peculiarity. It's still better than most Rust software, but logs and errors are cryptic. I'm yet to see a typical Rust stacktrace vomit, though I'm completely prepared for it.

The ugly:
Since there is no centralised authority, there are no centralised identities. Every node is represented by a public key. Which means, every one of your computers will have separate identity. While you technically can share keys between them, this isn't advised. This ultimately results in requiring some form of key management system, which I'm yet to explore.

Private repo support - while being there - is somewhat lacking. Someone with delegate access must list all nodes allowed to receive the repo, including your seed node. In my case, private repos require just three nodes for me alone. For a group larger than one person this might just turn into a nightmare. Have you ever managed SSH access with public key authentication? Similar story.

Seed nodes can either seed everything they touch or they can seed a select list of repos. There is no in-between, i.e. follow a select group of nodes and seed their repos only. Or at least, I couldn't find this feature. Which means, whenever you create a new repo and want to share it between devices using your seed node, you must SSH into it and manually add it to the list.

Discoverability is almost non-existent. Someone needs to provide you with a hash for repo to clone before you can work on it. Some seed nodes employ a web interface to list repos and browse code, but it's less than ideal. Same goes for discovery peers.

@monsieuricon by the way, thanks a lot for writing b4, I feel like it lowers the psychological barrier to sending patches a lot for me

Do you want to know why #Musk, #Google, #Bezos, #Zuckerberg are backing #Trump?

Easy.

The #Biden administration, in policies likely to be continued by the #Harris administration, have been enforcing anti-trust regulations at a level not seen since the 60's.

#Harris2024 is viewed as an existential threat to their profits. How can they engage in #Rent seeking behavior?

Yes, they are happy to have #fascism if it means continued #profits.

KnowBe4 hired a software engineer. As soon as they received their laptop the SOC light up like a christmas tree because of the malware it was loading up.

Working with Mandian and the FBI, it turned out it was a fake IT worker from N. Korea.

https://blog.knowbe4.com/how-a-north-korean-fake-it-worker-tried-to-infiltrate-us

Yeah, https://cdn.redhat.com/ returning a TLS error is exactly what I needed right now.

I'm not an expert on the "electoral college" system, but I believe this is when Russian troll farms are trying to make sure Trump wins, while Chinese troll farms are trying to make sure that Trump loses.

Time for the Internet’s many neurology experts to find new fields.

what if the crowdstrike thing wasn't incompetence but a failed backdoor attempt

Rarely have I nodded this much when reading an analysis of the systematic failure of the tech industry and our "market system" as a whole.
The unending need to cut cost (at all cost) to improve "metrics" that have nothing to do with creating sustainable, quality results... Yeah, it's bleak. It's terrifying.
https://www.wheresyoured.at/crowdstruck-2/?ref=ed-zitrons-wheres-your-ed-at-newsletter

Millions of linux users around the world had their productivity crippled by #cloudstrike as they all rushed to read the news, post hot takes & dad jokes to their Mastodon feed.

Kids these days! Back in '95, we got a Windows Blue Screen of Death 5-10 times a day, and that was on a *good* day.

Reminds me of that time when, a good few years ago, I woke up to a flurry of monitoring alerts reporting that every single kernel.org system went offline. After hours of scrambling, I managed to track it down to an automatically applied selinux package update that depended on another update in rsyslog that wasn't marked as "security" and so wasn't applied in the same transaction. This caused rsyslog to wedge, and if syslog calls are not returning, that puts the entire system into a massive gridlock.

Anyway, that was the end of automatic OS updates on all critical path systems.

Fucking wonderful. I get COVID on a work trip, then get laid off while still recovering.

Who needs an SRE? AWS, GCP, all the usual stuff you'd expect. Immediate start, obviously. US-based, either remote or a company that will pay relocation to a blue state (any non-swing-state is fine, other than VA or NH)

DM for CV. Tell me who you work for first for the version with unredacted history, or just DM a listing for me to apply to, let me know if you want to give you as a referral or not. If you have a redacted version I will of course provide a full one directly to the company on request.

Standard stuff: Have green card. Absolutely can not get a security clearance.

Edit: Thanks so much for all the support so far ❤️

#GetFediHired

Someone, non-ironically, sent me a gopher:// link today. 👴

Inbox 90 is the best I can do today. Considering I started with over 3,100, this is not too bad. :)

Not the mental image I needed today.

---

To: admin@kernel.org
Subject: current state of recent kernels

Why do recent kernels require so much twerking, to execute Suspend and
Hibernate Successfully? If you are Lucky!!

I am amused that "sovereign" now means some kind of cross between "independent" and "self-sustaining" when it literally means "overlord."