In civilized nations, "let the buyer beware" was discarded when we discovered that it led to higher market efficiency to have consumer protections in place, allowing most consumer to trust the information in the labels and ads.

We are discussing if some speech may be restricted in spite of the general protections of speech.

The case of yelling fire in a crowded theater is an example, showing that there are other principles that may override the protections of free speech.

Consumer protection laws agains misleading advertisements are another example, quite relevant to the OP.

Laws against incitement to crime, and laws agains sedition, are other examples.

So, the freedom of speech is not absolute, and will likely never be. The issue then, is how to draw the line.

That's how buffer overflows are exploited, provided the buffer resides in the stack frame of a subroutine. The analysis program is buggy, and overflows the buffer with data to be analysed, Next, the buggy subroutine finishes and returns to the caller. But the return address has been overwritten too. If the data written over the return address has been carefully chosen to point to data in the overwritten area, the program will begin executing information contained in the DNA.

Thanks, that resolves it. Of course, comparing a proton to a carbon atom is very different from comparing it to a kilogram prototype.I failed to google and find a codata value for m_p/amu (mass of proton/atomic mass unit), and I did not think deeper about the uncertainty in the kg prototype.

Now the computation becomes p_m(codata) minus p_m(new measurement), compared to uncertainty in p_m(codata).
The difference is 296 x 10^-12,
the uncertainty is 91 x 10^-12,
the ratio 296/91 = 3.25.
The new value is more than three standard uncertainties less than the old one.

The PRL seems to be paywalled, but the codata 2014 value in atomic mass units appears in the en.wikipedia page for "proton".

Googling CODATA values:
proton mass = 1.672 621 898 (21) x 10^-27 kg
Atomic mass unit = 1.660 539 040 (20) x 10^-27 kg
Releative standard deviations: 1.25 x 10^-8

Ratio of codata values: 1.007 276 467 285 (i.e., codata proton mass in terms of atomic units)
New measurement: 1.007 276 466 583
Difference: 7.0198469259707963 x 10^-10
Relative difference: 6.9691362341583399 x 10^-10

How is this three standard deviations?

This resembles the rapist who thinks the girl forced him to do it by being so female and attractive.

The linux kernel was there first, GPL and all. Nobody was "forced" to write GRSecurity as a patch to Linux. Nobody wrote code innocently only to have it taken away from them afterwards. GRSecurity does not work without the Linux kernel, or, if you can make it work without, you are free to do so,

This is another distortion of the facts. The code does not "happen to link" totally by accident or by evil acts of the Linux crowd. First, I doubt it just links, without any patching of existing code. We are talking about applying patches, that is, creating a derivative work in the form of a modified compilation unit. Who is "taking" other people's code here? And who is applying the patches? Who is doing the linking?

And now, after the split, could you fix the perennial "the printer is offline" annoyance?

I mostly second this. But please stop confirming, even if only initially and partially, the idiocy of the fairness argument. If anyone earning $10^7 thinks it is "unfair" to pay 70% taxes on that income, he is wellcome to switch position and start flipping burgers instead.

An income of $10^7 is only possible through working the society. Leveraging the rules, taking advantage of empowering circumstances in society. You don't do that working your farm on an isolated island. But society is there for everybody, not just for the 1%. Burger flippers are poorly paid, not because it is fair, but because burger flippers have little leverage and little power. It is not a matter of protecting the human rights of the 1% to keep it that way. Using the voting rights and electing politicians that tax the wealth, is a reasonable way we burger flippers and other 99%ers can wield our power against the powers of the 1%.

Solar particle radiation knocks hydrogen out of the earth's gravitational field. That is, hydrogen atoms achieve escape velocity of 11.186 km/s or more. I guess they mean that the Earth and Moon remained sufficiently hot for sufficiently long.

I thought that Earth was already largely devoid of water since its creation. The creation of Earth amounted to a large mass (one earth mass) falling freely from large distances into the center of the cloud from which the Earth formed, and so hitting the proto-Earth at velocities near the escape velocity. The surface of Earth is thought to have cooled in a few hundred thousand years, but that was probably more than enough time to send all the water to the outer atmosphere and out to cosmos. And besides, most of the water had already disappeared from the dust cloud that gave rise to Earth before that cloud coalesced to Earth, thanks to solar radiation which probably began before Earth reached any appreciable size.

Earth regained water after cooling through comet bombardment. Comets formed sufficiently far away from the sun to be able to keep its water.

VM's have a different hardware signatures from the host, and the Windows media most likely (or in my case: certainly) won't run. And, if it runs, it won't be "genuine". Maybe the pro edition has a virtualization extension that allows you to run VMs, Those who know, please tell.

Would it be possible to establish additional trust mechanisms, like this?

Establish a service which crawls the internet weekly, and keeps a hash of all new certs seen. Let there be multiple such services run by independent groups. Let such services also keep track of certs that have been revoked.

Then modify an open-source browser to emit queries to one or more such services, asking if the hash of the cert in question is OK.

This allows the users to choose who they trust. It would detect most MITM attacks, as the MITM would present to the victim a cert that would not be known to the service, unless the MITM has previously MITM-attacked the service as well.

Of course, the browser should also keep it's own cache of known good certs. This would greatly reduce the load on such services.

The responses, if affirmative, should be like certs signed by the service. The queries would be encrypted to the service's key, and would contain a symmetric session key to use to encrypt the response.

As an alternative approach, the query could contain also the url being visited. If the service has never crawled this host, it could visit it now, and see if it got the same cert. This would be slower, but would make it work even if the service does not yet have the resources to crawl the entire net, or if the client is visiting an isolated node.

From the article:

“What the certificate does not give them the ability to do is issue public certificates to other organizations. That's the big misunderstanding.”

What does this mean? Could it be that they only can issue certificates for "*.bluecoat.com"?

If so, what is the problem?

If so, why scrub the net of the incident? Why not proudly promote the story?

I always wondered why there is no distinction between macros that only modify the document in which they are embedded, and all other macros. Say, for instance a letter template that, upon instantiation, sets today's date, then removes all macros from the document.

It's the margin that counts.