A look at Project Liberty and DSNP. The Emperor has no clothes.
The Digital Standard Networking Protocol (DSNP) and its ambitious backer, Frank McCourt, have recently made headlines with plans to acquire TikTok, but a closer look reveals a troubling reality. Despite years of development and tens of millions of dollars in funding, DSNP remains shockingly underdeveloped, embodying a classic case of vaporware. While the intentions behind DSNP and Project Liberty are commendable, their execution falls woefully short, marked by minimal progress and a lack of essential features, leaving many to question how so much effort and investment could yield so little.
The Curious Case of DSNP: Years of Effort, Minimal Output
It's been a while since I looked at DSNP. There’s news that its backer, Frank McCourt, is trying to put together a team to buy TikTok. That’s insane. The DSNP and Project Liberty team are well-meaning, but they’re even less equipped to run a social network or protocol than Elon Musk at his lowest point when he was firing engineers based on how many lines of code they’d written.
To check what’s happened with DSNP since I last looked a year ago, I just read the entire spec, and there's not actually that much to it. The amount of time and money spent on developing so little is quite stunning. While the spec is fine, it is minimal.
Key Features and Odd Choices
DSNP allows users to have a key pair, write content into a datastore, follow people, and sign messages. However, the protocol seems to mix up layers, specifying the wire protocol, signing system, and at-rest specification format that clients and user data servers might use (pds in atproto, instance in fedi, relay in nostr). They use the ActivityStreams 2.0 schema, which is a good choice. Many new protocols have created their own schemas, but I appreciate DSNP's use of the standard. I wish more protocols had not tried to reinvent the wheel in this respect.
Encrypted Direct Messages and Metadata Privacy
There’s mention in the whitepaper about encrypted direct messages, suggesting it’s possible to hide metadata using techniques from Ethereum’s Whisper. Although the specs or code about DMs don't elaborate much on this, it’s good to see an attempt to hide metadata. However, there’s nothing about double ratchet encryption, forward secrecy, or any advanced encryption techniques. So, overall, it feels quite basic.
Data Storage and Interesting Choices
For storing user data and posts, they're using Apache Parquet, a file format from Hadoop. This is probably the only neat or interesting aspect. I've been searching for a supported data format like Parquet for a while because I think it can facilitate privacy-preserving contact discovery, so it’s cool to see its inclusion here.
Minimal Data Types and Following Options
DSNP supports very minimal data types: post, follow, reaction, profile, and, oddly, location. You can follow publicly or privately, the latter being a neat feature that Scuttlebutt had but most of the current generation of emerging social protocols don't.
Lack of Comprehensive Features
There does not appear to be any client app or server code, which might not be needed. There is an SDK for some core functions, but that’s it. Their documentation lacks information on user discovery, algorithms, or anything beyond the basics.
The Verdict: A First Draft After Years of Work
It feels like someone spent millions and ended up with a naive first draft, despite working on this for years. It’s similar to the early sketches of Farcaster and Lens, but those protocols have grown significantly over the past couple of years, with real users, lots of functionality, and apps.
DSNP remains as vaporware today as it was four years ago. These clowns have wasted tens of millions of dollars on nothing. Some academics have happily taken the money to fund their research labs. Consultants have written and thrown away countless prototypes nobody’s ever used.
There are lots of serious projects to build decentralized social media protocols. All these folks had to do was either work with others or even copy them. This stuff is all open source. Yet DSNP keeps sending folks to conferences without talking to anybody or attending any talks.
It’s a shame because they’ve got their hearts in the right place. The Unfinished conferences are really well put together with lots of great speakers. They hired the supremely talented and savvy Karl Fogel and James Vasile to help them make their code and process open, but you can’t fix DSNP.
Fundamentally, software and specs are a social enterprise, even more so if you’re working on social protocols. But these folks… they’ve never figured out how to play well with others. The money and good intentions couldn’t overcome their fundamental failure to have the humility to learn and play well with others.