August 14, 2014
Heartbleed v Ethereum v Tezos: has the Open Source model utterly failed to secure the world's infrastructure? Or is there a missing trick here?
L.M. Goodman stated in a recent paper on Tezos:
"The heartbleed bug caused millions of dollars in damages."
To which I asked what the cites were. His immediate response (thanks!) was "Nothing very academic" but the links were very interesting in and of themselves.
First up, a number of the cost of Heartbleed:
....To put an actual number on it, given some historical precedence, I think $500 million is a good starting point [to the cost of Heartbleed].
So, read the entire article for your view, but I'll take the $500m as given for this post. It's a number, right? Then:
Big tech companies offer millions after Heartbleed crisis Thu, Apr 24 12:00 PM EDT By Jim FinkleBOSTON (Reuters) - The world's biggest technology companies are donating millions of dollars to fund improvements in open source programs like OpenSSL, the software whose "Heartbleed" bug has sent the computer industry into turmoil.
Amazon.com Inc, Cisco Systems Inc, Facebook Inc, Google Inc, IBM, Intel Corp and Microsoft Corp are among a dozen companies that have agreed to be founding members of a group known as Core Infrastructure Initiative. Each will donate $300,000 to the venture, which is recruiting more backers among technology companies as well as the financial services sector.
Other early supporters are Dell, Fujitsu Ltd NetApp Inc, Rackspace Hosting Inc and VMware Inc.
The industry is stepping up after the group of developers who volunteer to maintain OpenSSL revealed that they received donations averaging about $2,000 a year to support the project, whose code is used to secure two-thirds of the world's websites and is incorporated into products from many of the world's most profitable technology companies.
What is truly very outstanding is that last number: $2000 a year supports an infrastructure which the world's websites reside on.
Which infrastructure was hit by a minor glitch which caused $500m of costs.
This is a wtf moment! What can we conclude from this 250,000 to 1 ratio? Try these thoughts on for size:
- Open source drives the SSL business because Apache, Chrome and Mozilla control the lions share of activity in SSL. Has the model of open source failed to keep ecommerce reasonably secured? What appears clearer is that the open source model adds nothing to the accounting for the value to society of this infrastructure. We could argue that accounting isn't its job, but actually some proponents argue vociferously that source code should not be charged for, which is an accounting statement. So I'd say this is a germane point, because the marketing of the open source community may be making us less secure if OpenSSL developers find it hard to charge for their work.
- The "many eyeballs" theory is open source's main claim to security. Is this a sick joke which just cost society $500m or is this an outlier never to be repeated? Or proof that it's working?
- This all isn't to say that the paid model is better, the paid alternative includes its disasters. But the paid model does typically carry liability and allocate maintenance out of the revenues. Open source doesn't seem to do that.
- Echoes of Y2K -- even though the combined spend was $500m, we still see no damages. No bad guys slipped in and stole any money, that we know of. Yes, there was one attack on CRA which cost a few hundred data sets, but again because the damage was caught before, we simply don't know whether spending $500m saved us anything.
-
The direct cause of costs here is one of upgrade. A sysadm wants to hit the button, and upgrade from BAD OpenSSL to GOOD. Why is that so hard? How do you upgrade SSL? Fixing bugs works in slow time because of burdensome commit privileges and the long supply chain, putting through protocol changes works in even slower times. At the protocol level, the IETF working group process is good at adding in algorithms (around 350 available, yoohoo!) but has no answer for taking things away; the combined effect of these 'essential processes' leads to an OODA cycle of 3.5 years to 80% rollout, as measured over the renegotiation bug.
This is not an attack on the people, and the ones I've met are not bad people, diligently doing their part. This is an attack on the change process, which sucks, today at a power of 250,000 to one.
$500,000,000 ⇒ $5,000,000 → $2,000
This is a widespread, burning issue, so let's look at two positive lessons from the Bitcoin world.
Bitcoin faces the same developer shortage. As Bitcoin developers get snapped up by well-heeled startup ventures with millions in VC money, and as the altCoins and side-chains and ripples and ethereums and now Tezos snap at heels with alternatives, the need for change goes up while the developer availability goes down. L.M. Goodman which makes the same point that upgrade is the archilles heel of all successful software systems:
Abstract: The popularization of Bitcoin, a decentralized crypto-currency has inspired the production of several alternative, or "alt", currencies. Ethereum, CryptoNote, and Zerocash all represent unique contributions to the crypto-currency space. Although most alt currencies harbor their own source of innovation, they have no means of adopting the innovations of other currencies which may succeed them.
Is this the same thing that happened to OpenSSL?
As an emerging model, new startups such as Ripple and Ethereum have done pre-mines: massive creation of paper value before letting loose the system in the wild. These paper values are then hoarded in foundations in order to pay for developers. As the system becomes popular, the value rises and more developers can be paid for.
Now, leaving aside the obvious problems of self-enrichment and bubble-blowing, it is at least a way to address the problems highlighted by the Heartbleed response above. For example, last Friday, Gavin Woods stated that Ethereum had raised $15m or so in BTC before they'd even shipped a real money client, which puts them several times ahead of OpenSSL. Not shabby, especially compared to the combined efforts of the world's powerful tech cabal.
And, stupidly thousands of times ahead of OpenSSL's contributions pittance ot $2000 per year.
Of course, this situation only applies to a very cool segment of the market: those cryptocurrencies which manage to garner mass attention. But it does raise a theoretical possibility at least: imagine if every open source project were also to issue their own currency?
And do their pre-mine, with say 50% reserved for developers? Obviously, it's valueless stuff at the start ... until the project booms in popularity, and the currency rises in value. Which is the alignment we want -- cash for programmers as the software starts to prove itself.
Think about a new model of open source + foundation + pre-mine -- if OpenSSL or Eclipse or Firefox were their own money, they'd also solve the problem of paying for developers. (The obvious problem of "Eclipse is not a currency" is just your problem in experience, contact any experienced financial cryptographer for how to solve that.)
Then, once you've got the money, how does it get spent? Upgrade is also a huge problem for the Bitcoin world. Adam Back has proposed two-way pegging to address the need to set up side chains for development purposes and also altCoin purposes. I've heard other ideas too, and for once, Microsoft and Apple are on the right side here with their patch Tuesdays and App Store processes.
Close with Goodman again:
We aim to remedy the potential for atrophied evolution in the crypto-currency space by presenting Tezos, a generic and self-amending crypto-ledger. Tezos can instanciate any blockchain based protocol. Its seed protocol specifies a procedure for stakeholders to approve amendments to the protocol, including amendments to the amendment procedure itself. Upgrades to Tezos are staged through a testing environment to allow stakeholders to recall potentially problematic amendments.
Maybe the new model is open source + foundation + pre-mine + dynamic upgrade?