back
16 comments
The picture this article is painting is almost completely wrong on every point that matters.

First of all, the NSA has not sponsored development of Accumulo for more than a year. The gist of the story is that companies that charge to support other implementations of Big Table-like systems argued that the government support of Accumulo directly interfered with their ability to compete. Their main point was that the government's job was not to compete in an already-thriving industry, and their lobbying was successful: in summer 2012, the NSA was formally ordered to halt production on Accumulo. Some of the employees left to start their own company, last I heard.

Second, as other users have noted, this is not the only implementation, and dropping it will certainly not make it impossible to do what they do. Remember that Accumulo was started when no OSS technology could do what it did. Now is a different time, and arguably it is actually behind.

I agree that it feels good to stick it to "the man", but let's be realistic here. This move is accomplishing nothing.

Accumulo is designed deeply with authorization and access control in mind. Every time Ive spoken with [edit: people at the] NSA, I wind up hearing about how much time they spend just jumping through hoops to avoid looking at information they aren't supposed to.

Accumulo is also a suitable data store for education and medical data, where there's also lots of regulation over who can see what.

The linked article is reactionary and has no substance.

"I wind up hearing about how much time they spend just jumping through hoops to avoid looking at information they aren't supposed to."

If they didn't collect so much, it'd be easier to avoid the info they shouldn't be looking at.

Even if they “didn’t collect that much” they still need to protect sensitive data.
Exactly. Medical records are a great example. If you get a new doctor, you have to give them written authorization so that they can get your old medical records.

They can't just say "look I'm a MD, give me the datas", partly because that would be a felony!

Same sort of regulation, albeit more complicated, applies to NSA.

This is a political approach--"I don't like one aspect of what an organization does, so I will harass anyone who associates with them in any way."

It's pointless in this case because the NSA is not subject to such grassroots political pressure. Their sigintel teams do not give a shit if you use Accumulo or if Apache hosts it. Harassing Apache will have zero effect on Prism and related programs, but will harm Apache, which is a known-good organization.

If you want to change things at the NSA, the political pressure points are in Congress and the White House. Not Apache.

It's exactly Apache's known-good status that makes it a point of leverage. Surely the pragmatic thing to do is to put pressure on organizations that you think may respond, rather than organizations that you know will not. I really don't understand why everyone is saying that Apache's hosting of this project is irrelevant and useless to the NSA. Would you say the same about other projects that Apache hosts?
The NSA is one of the most well-funded tech organizations in the world. Apache does nothing for them and has no leverage on them whatsoever.
As some of the comments on the linked article point out, Apache Accumulo is one of several implementations of BigTable; another one, also hosted by Apache, is HBase. As far as the fundamental technology goes, this particular cat (key-value stores scalable to very large data sets) is out of the bag, and used by many people besides the NSA.

Would Apache dropping support for one particular BigTable implementation meaningfully hinder the NSA?

"Open to all contributions, as long as we agree with you" is a pretty slippery ethos.
So we want them to be more open but we don't want to host their open source code for all to see? Believe me, I'm sharpening my pitchfork and lining up my torches over this thing, but that just doesn't make any sense.
mpyne. I expressed it badly - I did not mean to say that open source developers are supplying the NSA with Accumulo, just that Apache is supplying the NSA with a useful service by hosting Accumulo as a project (and that for some reason open source developers are OK with this in a way that they would not be for private corporations selling closed source software to the NSA).

The general picture seems to be what antics says in the first comment: NSA adapted Big Data technologies to suit their own access control and other requirements, and then decided that the best way forward was collaborative. A group of employees set up sqrrl to provide a commercial distribution. But it's an ecosystem that the NSA gains from, all the same, and the broad membership of Apache seems like it makes for a place where people may have leverage.

As the poster of the linked article, I'd like to respond.

antics' "First of all" paragraph is completely consistent with what I wrote. The "Second" paragraph is an "If I don't do it someone else will" argument: it's not acceptable for arms companies and I don't see why it's more acceptable for open source foundations.

cartershonwald's first paragraph is also consistent with what I wrote. I am sure that the NSA takes more concern over privacy than, say, Google or Yahoo. But that's not really relevant here is it? And the fact that the technology is also useful for other things - well, we can say the same for internet filtering technologies and deep packet inspection, but we still criticize Blue Coat and others when the sell that to authoritarian regimes. I believe what I'm saying is consistent with criticising those companies.

_delirium's comment is also "If not Apache, someone else will do it" as above. Not good enough for arms manufacturers, and not good enough for the Apache Foundation.

I don't understand showerst's comment.

Pherdnut - I don't think the main argument with NSA is that they are being secret, it's that they are engaging in massive and illegitimate surveillance.

For showerst's comment, he's basically saying that the idea of e.g. having a hardware store refuse to stock a screwdriver just because it was invented by a nuclear weapons development lab is a pretty slippery slope ethically.

After that we won't stock screwdrivers invented by the government (incl. EPA).

Then we won't stock screwdrivers invented by $UNDESIRABLE_GROUP.

That line of reasoning is why both the FSF and DFSG explicitly say that Free software cannot have terms-of-use restrictions. If it's free for one to use it should be free for all to use, even those you hate.

Accumulo is a tool. It's a tool that happened to be useful for NSA, but NSA is not the only group that has "Big Data" problems and so it makes no sense for Apache to punish the rest of us because Accumulo has a source that you personally don't like.

mpyne - thanks for clarifying: I see it now. I've never been a fan of "slippery slope" arguments because once you start on them, you never know where they finish.

I know why FSF takes the line it does, I'm just saying it is a line that has taken us to a strange place where open source advocates are simultaneously supplying the NSA with services to carry out surveillance, defending the practice, but complaining about the surveillance.

And as for "Accumulo is a tool". That's kind of a "guns don't kill people" argument. Would you apply the same argument to private companies supplying authoritarian regimes with deep packet inspection software? I see little difference in the cases.

I'm not a super big fan of slippery slope arguments either. Then you get people opposing programs not on their own merits, but what might happen if that program were in the hands of the "other party", and then later in the hands of some dystopian totalitarian state.

Either way, I'm not sure why you say open source developers are supplying the NSA with Accumulo. You have it completely the wrong way around; NSA supplied us with Accumulo.