The Google Fonts case was decided based on the transmission of the full IP address in a jurisdiction (Germany) where there are ways to identify a user by means of that address. CNIL's press release follows a decision by the Austrian data protection authority where the Google Analytics cookies were at issue.
If you can read German, you can look at the Austrian decision directly, the complainant has uploaded it at [1] and the relevant section is D.2 b) starting at page 27.
Strictly necessary cookies for a service the user explicitly requested. And, importantly, this is true even if no personal data is involved and the process is therefore not covered by GDPR at all -- the cookie clause of e-Privacy Directive applies regardless.
You're making the mistake of thinking that the cookie consent requirements are somehow a consequence of GDPR. The cookie consent requirements exist separately from and additionally to GDPR as a consequence of the e-Privacy Directive. What GDPR changed in regard to cookie consent is what exactly constitutes "consent", as it updated the Data Protection Directive in that regard, but it did not change when consent for cookies is required.
Other than court judgments, the Article 29 Working Party opinion is the most authoritative opinion you will get on the interpretation of the e-Privacy Directive, which is the "real legislation" that you need to look at.
edit: Nobody claims that the e-Privacy Regulation is in effect, by the way -- of course it isn't, it hasn't even been passed. The cookie consent clause of the e-Privacy Directive is however in effect, and has been since 2009.
It is not true that "functional" cookies are generally exempt from the consent requirement. What is concretely exempt are necessary cookies for a service that the user explicitly requested. This is not the case for cookies placed by Instagram embeds.
The website tried to rely on legitimate interest as the legal basis for processing the data, and that precisely requires a balancing test between the interests of the website host and the interests of the data subject.
If you want to make sure that you're not getting the balancing test wrong, you can always go for the legal basis of last resort: consent. Just ask the user whether you can load content from Instagram and only do it if they agree. In fact, since in parallel to the question of your legal basis under GDPR, you also have to comply with the cookie provision from the e-Privacy Directive, where there is no "legitimate interest" exception to the requirement to ask for consent, you will have to ask for consent anyway (as Instagram embeds place cookies).
A MAC of a message m can only be computed with the knowledge of a key K. Specifically, with a cryptographic hash function h,
HMAC(K, m) = h(K + a || h(K + b || m)),
where + is addition mod 2 (xor), || is concatenation and a and b are constants. (This construction takes into account possible length extension attacks on h.)
Given that h is secure, knowledge of any reasonable number of pairs (m, HMAC(K, m)) does not allow you to recover K, and without K, you cannot compute HMAC(K, m) for known m, i.e. enumerate all the possible MACs for serial numbers.
p^μ is the μ-th component of the vector p, and in an equation p^μ = m u^μ, μ is to be taken as a free variable, i.e. the equation is true for every μ. In relativity, Greek indices are taken to range over time and the three spacial dimensions (whereas Latin indices only range over the spacial dimensions).
This notation can be naturally extended to tensor products of vectors in the tangential and co-tangential spaces to the base manifold that is spacetime (simply called "tensors" by physicists): https://en.wikipedia.org/wiki/Einstein_notation
It may be forgivable to use dynamic mass in an article written for laymen and using a simple model of the atom, but I feel I should mention that the concept of objects changing mass depending on the reference frame is a very dangerous one because substituting the dynamic mass for the mass in a classical formula does not always lead to correct results.
The concept of dynamic mass is motivated by wanting to continue to write the previously known three-momentum as p = m v, which does not conform to special relativity, hence the definition of mass is changed. However, in a formula as basic as F = m a (F and a being vectors), substituting the dynamic mass for m does not yield correct results because in general, under special relativity, F and a do not even have to be parallel.
Modern formulations of dynamics in special relativity use the more intuitive invariant mass, and three-momentum is written as p = m gamma v, where gamma is the factor previously included in m_r. This p is now the spacial components of four-momentum p^\mu = m u^\mu, where m is the invariant mass and u is the relativistic four-velocity of the moving object.
This is what 23andMe said about that person's genome:
"Has two mutations linked to limb-girdle muscular dystrophy. A person with two of these mutations typically has limb-girdle muscular dystrophy."
Of course this not a diagnosis in a technical sense, but to a lot of people it will sound like it is. Specifically, I don't see how this is "very clearly" not a diagnosis.
This reminds me of a blog post in German [1] by a person who due to a software bug had been falsely diagnosed by 23andMe with limb-girdle muscular dystrophy. (Fortunately, he was able to identify that it was a misdiagnosis.) It appears that in this case potential misdiagnoses aren't just a theoretical problem.
Obviously Assange (et al.) didn't ask anyone if they could leak sensitive data -- and perhaps they should have (but how exactly would that have gone do you expect?) -- but I believe Assange is acting on this above principle.
In summary, he treats government as a conspiracy (liberally defined to be a social network) whose total conspiratorial power (the sum of the weights of the edges of the social graph, where the nodes represent people and the edges the potential to share information) can be kept in check by increasing the cost of exchanging information -- which is exactly the effect of leaking classified material.
It's quite silly that the author doesn't know this, but the feature he requests exists already and has existed for 10 years -- Privacy Extesions for SLAAC in IPv6 (RFC3041). It's even enabled by default on Windows.
If you can read German, you can look at the Austrian decision directly, the complainant has uploaded it at [1] and the relevant section is D.2 b) starting at page 27.
[1] https://noyb.eu/sites/default/files/2022-01/E-DSB%20-%20Goog...