Hacker News Reader

Comment by ForHackernews | Hacker News Reader

Why do you believe that a JavaScript port of bitcoin is an important project? What is lacking in current implementations / why would you want to run bitcoin inside a browser / do you think it's possible to do serious crypto correctly in JS[0]?

[0] http://matasano.com/articles/javascript-cryptography/

View on HN

There is no working javascript full node. I believe the future of bitcoin is brighter if there are many independent implementations of bitcoin running as full nodes.

I want bitcoin running in a browser because browsers work on every platform and can bring bitcoin to more people. Also keep in mind that reddit is primarily a website, to using javascript is a pretty appropriate first choice of language for bitcoin stuff.

Can you do serious crypto in C? Only if you know what you're doing and don't screw up. Same goes for javascript.

>Can you do serious crypto in C? Only if you know what you're doing and don't screw up. Same goes for javascript.

Overall I am sympathetic to your cause, but this raises alarms. JavaScript crypto is nonsense, you could have the implementation silently replaced with a naughty one at the hosts leisure.

EDIT: To respond to all of you at once - everything I've seen so far says that he was aiming for JS in the browser. Did I miss something?

That's not really the case - yes, crypto in the browser is possibly nonsense, but crypto under e.g. Node.js is anything but.

We're not talking about client-side javascript, are we? I assumed he meant node.js/server side.

Yes exactly. I think people who are not whatsoever involved in the javascript world assume that javscript means "delivered by a webserver", which is just not necessarily true.

For what it's worth, I (the acuser) am super familiar with JavaScript, I just thought you were making a web app. Sorry to cause confusion.

> JavaScript crypto is nonsense

What makes you think this?

That's also true of C with an auto-update framework. That's also true of closed-source C.

There are plenty of ways to avoid that particular trap of JS crypto (or more specifically web crypto). Bundling it inside an extension comes to mind.

All that said, I'm not even convinced this attack is within the threat model of Redditcoin, given that it's resources for a specific website that is also the JS origin.

That has nothing to do with javascript-the-language, but rather the way the web works. Obviously even with https, javascript delivered by a web server has a vulnerability whereby the web server may be compromised, which is outside of the users control. Any javscript delivered by a web server should be executed with that understanding. i.e., delivering a wallet from a web server is probably not the right thing to do. Note that javascript can also be run in the form of a browser extension, server-side node, node-webkit, apache cordova, etc. Javascript-the-language is not intrinsically vulnerable to a compromised webserver.

It's fair to say that JavaScript-the-language is not intrinsically vulnerable to this, but how else are you planning on delivering the JavaScript to clients? I may be misinterpreting your goals.

Not in the form of a webapp. It would have been a browser extension and/or mobile.

Thanks for clarifying.

This seems pretty much correct.

A usable browser implementation of a wallet could be secured fairly well and if it gained any popularity would put pressure on browser makers to deliver better solutions.

npm can use server-side resources.

But where do you store the 25GB blockchain in a client-side full-node implementation?

Isn't this just an implementation only for Node?

I don't think anyone is going to run a full node in a browser any time soon, but if they do, IndexedDB.

Bitcoin's cryptologic is open-source; it doesn't have to be secure from tampering in the same way a videogame console has to be secure. A videogame console doesn't want you modifying game code or running your own code on the hardware. Nobody cares what you do to your bitcoin client; you'll just lose your coins in some invalid[1] transaction or maybe you find some amazing flaw in bitcoin's math - which you could have found in any other programming language anyway.

That said, a browser-based bitcoin client does expose its user to potential XSS exploits 'n such; especially if the private key is in the browser's memory.

1. https://en.bitcoin.it/wiki/Raw_Transactions , basically if you mess this stuff up you lose coins.

That's a fair point, except that (contrary to "Nobody cares what you do to your bitcoin client"), I would suggest that most people do not have their own bitcoin client, and are in fact running code written by somebody else. Those users will likely be very upset if a client loses all their coins.

Or, if there is a large enough user base running some faulty logic, this could cause a big ugly fork in the blockchain. Not good.

Yes, that is the primary concern. "We" in the bitcoin community haven't yet figured out how to write 100% correct alternative implementations of bitcoin. We will have to figure out how to do that first before I would ever recommend anybody actually run an alternative implementation as a full node and rely on it for seeing the blockchain.

Reply on news.ycombinator.com