Hacker News new | past | comments | ask | show | jobs | submit | jgoodwin's comments login

I would play up the extension as an accessibility enhancement, and cite the ADA and accessibility guidelines. This makes the point about the users' 'right to transform data' more clearly -- access and transformation is about usability, and usability is a legal right.

The administration will greatly fear an ADA action, or arguments that come even close to them, while drawing parallels. Students crying 'Freedom!' is old hat, and will get a less effective response.


A bit off topic, but most of statistics also breaks.

If you go back to _Mathematical Statistics_ by RA Fisher, early in the last century, and look at his arguments about binning 'big data' into histograms, he has a nice little construction that uses the notion of an 'angle' running through the data set, does a Fourier Series expansion, keeps the 'DC' term from the cosine series, and waves his hand about second order effects. He does estimate them for the sine-like series, and finds for a data set of size N=1 Trillion it might be a 10% effect.

The only remnant of this whole proceeding in modern lore (and even Ph.D. statisticians may not have heard of it) is Sheppard's correction for equal class-interval histograms:

http://mathworld.wolfram.com/SheppardsCorrection.html

But of course when your datasets start to be 1 billion rows routinely, 10% effects a mere 3 orders of magnitude away in the size of the dataset should start to make you nervous.

Moral: once you get a billion data points of anything or so, it's time to redo the Maths, very very carefully.


I am a PhD statistician and I can't understand your comment.

What statistic is being calculated for the N=1 Trillion dataset? And what is the way of calculating that would be off by 10%?


I think there is a mathematical theory here people should become acquainted with -- differential privacy. So far it has found use in the context of a large data set, e.g. search engine query logs, to try to determine how invasive a statistical summary or release of only partly obfuscated data would be.

Like calculus, it has a sort of epsilon/delta construct -- given a differential privacy concern epsilon, under what circumstances (how tight a bound on delta) do I need to prevent that.

http://en.wikipedia.org/wiki/Differential_privacy

Perhaps this theory could quantify the intuition that while it's ok to snap my license plate now and then (very little differential privacy loss on my part), enough times and it becomes invasive, and dreadfully so.

This puts some teeth into the vague talk about a mosaic theory. The key idea is whether the aggregate information in the data set can triangulate you, to within say 10000 persons (not much privacy invation), or 100 (quite a bit). There seems to be a tipping point around clusters of 100-1000 persons or so, that is the typical size of small organizations or groups of people, such as churches and schools.

Now, as an application: license plates are nearly unique identifiers and the attacker has a database of who's who for all intents and purposes, so it is little different from asking everyone for their ID just because they are on the street. That's illegal by the way, even for the police.

The argument that was formerly deployed here is that using a vehicle on public roadways was a privilege that cost you natural expectation of privacy (non-intrusion) in public places.

The real crux of the issue here is that the public-private tradeoff was once predicated on the individual (the individual atom has protection, because data collection is sparse, so tagging the individual but not the path was meaningful). Now the data collection is dense, and even single particle tracks become visible.

If you think in terms of fluid mechanics, there's a sort of Euler view / Lagrange view here (as there is with tagged dollar particles and tagged wallets or accounts).

That is, tracking individuals and tracking their paths become duals of each other, if the data collection is dense enough. It doesn't matter whether the item tracked is the tagged individual, or the flows and transactions -- either way, complete reconstruction of the system becomes possible.

With any data set, there is a sort of 'phase transition' in its size, where you suddenly can see the underlying trajectories of all the tagged particles. Things that made perfect sense when data collection was sparse, just as allowing the police to jot down you license number and chase you with a bicycle, turn into totalitarian surveillance when the observations become dense enough -- in a way we can quantify in terms of a sudden jump in information gain that goes from nearly complete ignorance of where people are and what they are doing (the former phase), to near complete knowledge of everything. Very much like percolation theory.


asking everyone for their ID just because they are on the street. That's illegal by the way, even for the police.

That is not illegal at all. Anyone can ask for ID, including the police. Doesn't mean you have to give it to them. In the case of the police, refusing to show your ID can't be used as cause to arrest you. But they can still ask for it.


In a lot of countries it's certainly grounds for arrest as carrying a government issued ID is mandatory.


One of the very nicest things about the ReactOS project is their build chain -- you can download one project and have a very nice toolchain for building open source software targeting the windows platform, with very little fuss.

Consider the pain of getting a functional Cygwin system able to target non-GNU libraries.

Also, it's just plain fun to use the ReactOS desktop to look at your own desktop looking back at ReactOS ... and drag the ReactOS window down so it does the whole infinite mirror thing.


I haven't used a Mac since a bit after the introduction of quadro or whatever it was, so I don't have a dog in most of the discussion above. I also don't own a mobile device. Those facts probably date me as a Linux user.

Here's a suggestion for designers: the age of the Linux user base is aging at a rate of one year per year. That doesn't mean there are no young ones, only that we are going where no Linux user has gone before (I joined various Linux projects back in 1992.)

Every human on earth, after there mid 40s, experiences changes in the geometric optics of their vision system. As my eye doctor says, the denial rate is 100% and the participation rate is 100% -- if you are lucky enough to reach that age.

Please design an accessible operating system we can still use in our 80s. This is a very big opportunity. How old will you be in 2038 is maybe a good design target. ;)

Accessible on a mobile device, a desktop, and able to function smoothly with presbyteropia is the basic design spec.

Get your team an old geek geezer or two to tell what works and what doesn't. When I forget my reading glasses at work, I have to use the Magnify app on Win7. Go ahead and try this, and see what the experience is like and if it's so perfect it can't be improved on you are done. Otherwise, work on that aspect of your design.

Even if you are young and not 'there' yet, try out the Magnify app, and try whatever Linux has to offer on your distro of choice for the same thing. It completely changes how you size windows and operated effectively. It will give you insight into important principles of design such as navigating windows, scrolling, keyboard usage, etc.

There is no need to make the very best of what exists today -- make the very best that will be needed tomorrow.

You will appreciate this advice one day.


One thing that concerns me is that many Americans are adopting what is essentially a Tory/Loyalist attitude towards these events, without understanding the Tory/Loyalist political philosophy of dissidence, which differs from 'civil disobedience' and 'non-violent resistance' (those are the other guys).

Don't get me wrong -- we can't all be Patriots and Revolutionaries -- but our history has ill-fitted us to be good at being Tories and Loyalists. Those were the bad guys in all our grade-school stories ... and now we are those bad guys.

The classical Tory theory of dissidence is called "Passive Obedience." This doesn't mean bending over and being a wimp. It means being obedient to higher authority (God and Constitutional Law), while seemingly disobeying usurpers and tyrants, who are themselves violating the higher Law -- constitutional, moral, and natural. The "Passive" part is an old word meaning suffering (like the Passion of Christ).

Edward Snowden has given us a very good example of Passive Obedience -- if he is correct the programs are indeed unconstitutional. He certainly is suffering for his beliefs, and is fleeing, not resisting or rebelling against the State. Failure to obey the commands of usurpers and tyrants, or to obey bad law in defiance of the dictates of one's conscience, are not required even of Loyalists and Tories.

The fact that Tories and Loyalists, which the American people have become, are condemning his actions, shows only that we have forgotten how to be good Redcoats, as well as most certainly having forgotten how to be good Patriots.

As good Tories (not good Patriots though), Loyal to the American State, we have the right to petition our sovereign -- the American People, not its representative Government -- i.e., to request a constitutional convention to strike down these Star Chamber courts, redress the alleged tyranny, and end the usurpations against our Sovereign's previously constituted declarations, and granted Bill of Rights.

As far as Snowden's flight is concerned, Sir Thomas Hobbes gives a very clear explanation of both Passive Obedience and the right of the dissident to flee, in an attempt to evade the sure punishment he would otherwise receive with or without justice (however if he is caught he must meekly accept his Passion and martyrdom, without resistance -- Civil Disobedience and Resistance are the contrary of the Tory doctrine).

Time to pick sides -- but if we are going to be Tories all, let us not be bad ones. These are the times that try men's souls.


Someone should verify the 'large entity' discovered by Shamir is distinct: http://eprint.iacr.org/2012/584


I have both bitcoin and dwolla, and between the two I've found Bitcoin much more useful. Dwolla has a difficult business model these days, and I will be sorry to see them fail. But they will fail.


Perhaps. However the standard deviation for a Bernoulli trial is sqrt(Npq) where p = 0.00049 and q = 1 - p.

The two trials are not significantly different.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: