Bedrock – A modular, WAN-replicated, blockchain-based database | Hacker News Reader

Bedrock – A modular, WAN-replicated, blockchain-based database

bedrockdb.com

73 pointsdhruvkar25 comments

The "blockchain-based" part is misleading.

For one, when people say "blockchain", they're almost always including the clever protocol that enables a currency with no central authority. That's the interesting part. Bedrock doesn't do any of that.

Without that, "blockchain" is just a simple technique for incremental hashing. Bedrock uses that, but it's not substantial enough to sensibly list "blockchain-based" as one of the three top attributes.

(They could have said "Paxos-based". That's the protocol they use to ensure things don't get out of sync.)

Just trying to capitalize on the cryptocurrency hype, I guess. FaunaDB did something similar: https://fauna.com/blog/distributed-ledger-without-the-blockc...

just curious. Would you consider git as blockchain?

Hmm... Git's commit chain might be a blockchain (in the minimal sense of the word), since I think it uses that hash chaining technique.

But if I were describing Git to others, I wouldn't say "Git is a distributed, blockchain-based VCS" because that would give the wrong idea.

Not your op but I personally do not. The way the word has evolved blockchain includes the distributed consensus. I understand there is a concept of linked blocks, into a chain, and technically Bitcoin and folk are those. If someone said blockchain (datastructure) I would understand it as being only the data structure.

git is a chain of blocks, kinda. it's distributed. there is no consensus. It's not a blockchain by how most people would define it.

Not really. In bitcoin one chain is considered the truth, and this may change with chain length. in git, all branches are the truth.

I think it's a bit strange to call the basic idea behind a Merkle tree a "blockchain" -- we already had a word for Merkle trees before Bitcoin came about.

I do. My argument is here: https://www.fossil-scm.org/home/doc/trunk/www/blockchain.md

blocktree, more than blockchain.

Aren't there some altcoins that have attempted to do a blocktree? I haven't paid much attention to them, other than reading about one a while back, but it seems like that wouldn't really go anywhere because once you do away with a single, unified, public view of truth you might as well be using a lightning network?

You are absolutely correct in that most of the people think of blockchain as something that includes distributed consensus mechanisms, but strictly speaking from a technical point of view blockchain and proof-of-[insert flavour] are two separate things, so IMHO their claim that is blockchain based is correct.

The word only has relevance when talking about some consensus. Otherwise you are just adding a checksum or signing with a private key

> The "blockchain-based" part is misleading.

I don't agree. Maybe for one who doesn't understand what a blockchain is and only has an association with crypto currencies. They are just using a private blockchain as a ledger for all commits to the database. It seems a bit far fetched to me to assume they chose this core technology to "capitalize on the cryptocurrency hype".

They use blockchain as a synonym of "distributed ledger" (which blockchain is not just.) However, a distributed merkle tree is most of the "block and chain" part of the bitcoin paper. Yet, like other people in this thread, I believe Blockchain has come to include distributed consensus or byzantine fault tolerance, proof of work, and public keys as identities. Just because the word itself wasnt defined in the Bitcoin whitepaper, doesnt mean it hasnt been defined after the fact.

They also claim to have used it two years before bitcoin existed.

https://bedrockdb.com/blockchain.html

The "block" in "blockchain" refers to a certain kind of a block. Strictly speaking it refers to the "block" that is defined in the Bitcoin white paper (which itself doesn't mention blockchain): https://bitcoin.org/bitcoin.pdf

Extending from this, "blockchain" may mean any structure that has the same properties as Nakamoto's "blockchain".

Otherwise, the structure is just a hash list, a hash chain or a Merkle tree.

One need only read the words "private blockchain" on the referenced page to interpret this as "not a consensus blockchain".

I agree it's misleading; people need to understand that the words "private blockchain" are meaningless in the context of what they probably expect Blockchain (the capital B being what is on the page, again) to mean.

I note that the link for "private blockchain" attempts to explain this. But no one is going to convince me they aren't using the term for marketing hype.

This looks promising. A typical blockchain discussion in my org (kind of averse to change with a few hundred database people, smart, but Microsoft stack) goes like this: Someone: “We need a private blockchain for C use case. We’re about to pay consulting firm X a gazillion dollars to implement EthiHyperBitLedget” Architect: “Silly bird, you just need a distributed database...” Someone:”Fine then how do I get one of those” Architect:”you just hire me and I’ll write a bunch of database stuff for you that is a one off spaghetti mess of sql server stuff” 75% of the time they buy that blockchain thing 25% of the time they try to spatchcock the sql thing So far we are 0 for 10 for stuff being useful.

If there were a nice distributed database with the features of blockchain I think this could merge my excited Someones with my fuddyduddy Architects.

Wouldn't you clone Stellar or IBMs stuff already if you wanted "corporate" blockchain, i.e. a group of nodes sort of not 100% trusting each other, just keeping an eye on eacher.

This is the most mature and out of the box solution I've come across: https://kaleido.io/

I can totally imagine how Someone would respond to it because it wouldn't automatically get him the Lambo he thinks Blockchain should bring. After a couple of hours, he would say something like: "this isn't a Real® blockchain. We need a proper one."

The Architect wouldn't be impressed by it too, because where is the fun if you can't spend 6 months creating spaghetti code?

Isn't https://alacris.io/ pertinent?

Any distributed database that doesn't advertise that it has hired Kyle Kingsbury (aphyr) and passed him applying Jepsen's new and ever-kinkier stress tests against it makes me instantly suspicious. Appeal to authority? Yes, but, aphyr has earned the title by breaking a whole lot of databases.

So does this partition the data out to different nodes or replicate the same data to each node?

"Bedrock was built by Expensify, and is a networking and distributed transaction layer built atop SQLite, the fastest, most reliable, and most widely distributed database in the world."

I didn't know SQLite was a distributed database.

I think it's supposed to mean that it's the most used and not that it is actually distributed.

I read is as "(most widely distributed) database".

The SQLite site phrases it more precisely: "Most Widely Deployed and Used Database Engine" https://www.sqlite.org/mostdeployed.html

As others have said, "distributed" in this context means "widely used". However, there is work to make SQLite be able to work as a distributed database (and this code is being run in production on LXD clusters)[1].

[1]: https://dqlite.io/

Not a blockchain.

Reply on news.ycombinator.com