Posts by fg_swe 1478 publicly visible posts • joined 20 Nov 2021
...by Socialism. See Cuba and Venezuela.
Oracle meanwhile provides 170 000 well-paid, solid jobs for their employees. These people can afford good apartments, good and plenty of food, health insurance and at least one yearly vaction.
It is always amazing how supposedly well-educated engineers and technicians can be stubborn 1di0ts.
SUN gave away Java, after having sunk serious money into its development. While sinking even more money into the continued maintenance and improvement of Java. They destroyed the Delphi business in that process, because many developers got accustomed to this socialist model of technology development.
Then SUN had to be rescued from bankruptcy by Oracle.
Oracle is a well-functioning business, which in turn provides 170 000 employees solid and well-paid jobs.
You can argue all day about the specifics of payment for software products, but "totally free" is not an economic model, it is Marxist Stupor. See Cuba, Venezuela where they suffer hunger now.
Cybernetic attackers will analyze patches in order to attack not yet updated systems.
Patches should be thoroughly tested by the authoring company.
Also, they should be ditributed/staged encrypted on all affected computers and only after that the key should be broadcast and patch actually applied.
LSM also uses kernel level code to control and intercept potentially ALL userspace-to-kernel calls.
Yes, some security things must be done in kernel mode. But with that comes extreme duty of diligence by the "plugin" author. A config file error must never generate a bad pointer, but simply ignoring said config file.
CrowdStrike has a very bad quality assurance in place. The government should fine them for neglicence.
A thorough study of computer history will reveal that C, Unix and Windows are by no means the only "basis" of systems programming.
Rather, C and Unix (and Windows) can be seen as the Hamburger of computing. Quick and dirty, unhealthy creators of the Cyber War Domain.
Algol was already a memory safe alternative to C and Unix in the 1970s. It contained a lot of solid European technology. The beef steak with hand made noodles and a proper salad, in the meals analogy.
Unlike C and Unix, Algol mainframes were not given away "for free" though.
Again, I strongly assume IBM only allows the ANN AI to perform "safe" transformations of the original Query Plan. Worst outcome is bad runtime, but still correct results.
I am also quite sure all the transformations done by "conventional heuristics" are proven to be safe.
After all DB/2 and Oracle run billions of bank accounts !
1.) If the AI is only allowed to perform "safe" transformations of the query plan, there is limited worry regarding AI "correctness". No sane software engineer would grant an AI "full authority query plan transformation". AIs are "mostly correct", not "perfect".
2.) ANNs might indeed be a new type of heuristic for query plan optimization. Much more "rule of thumb based" than "statistically, logically based". It will prolly work nicely on standard scenarios and will prolly fail badly on special, extreme cases.
1.) My understanding is that the query plan mangling will never change end result, but only runtime/resource consumption. AI can only be used for heuristics, similar to existing heuristics in traditional optimizers.
2.) I have seen large enterprises where they run convoluted queries and still wait for hours in queue to get results. The business questions are similar to "how many millions did we lose in winter, in the US, because a component was not that great ?". There exist corporations who really need large scale relational databases to better understand their core business. Much more than just a funny CMS.
Now, if these flimsy web 2.0/key-value/graph "databases" contain 1/1000th of DB/2 query optimizer technology I would be surprised.
Again, relational databases are modern day wonders. They continue to be improved by scientists and engineers, as they are still far from perfect. Most users, including software engineers, never really cared about this technology. They just assume "it works" and only complain when the optimizer fails on an especially convoluted SQL query.
SQL database are on par with the latest jet engines, 3nm semiconductors, metal 3D printing, gas chromatography !
To your wet computer of 100E9 Neurons and 100E13 Synapses there exists an "obvious" query plan. But for the dumb CPU there might be Billions of alternatives and each of them must be "rated"/"estimated" and then the best one chosen. In reality, the dumb CPU will use some sort of heuristic to boil the billions down to 10000 different plans. Works often, but not always, well.
The premise is that the AI is essentially a novel heuristic for find a good query plan. Any experience from the real world ?
SQL Database Servers are "as old" as Unix kernels such as Linux, MacOS X and later, iOS.
I am quite sure there are serious teams at IBM, Oracle, MSFT developing incremental improvements of their SQL engines. I would not be surprised to learn that serious percentages of code are not older than five years.
I fully disagree with this notion. Advanced SQL Database are some of the most advanced systems in existence. As complex as a modern operating system. Critical for almost ANY business operation from accounting to inventory to sales statistics to analytical processing etc.
Relational databases continue to lead the field of databases due to solid theory, good performance, good query plan optimizers and so on.
The "new" key-value databases are merely useful for building massive collections of cat picture sharing systems for teenagers.
So far I can only read generic nonsense in this forum.
What I can say:
1.) Query Plan Optimization is a Hard Problem. MySQL fails badly in this aspect for complex queries.
2.) For many queries it would be too runtime-expensive to evaluate all possible solutions; heuristics are used.
3.) Maybe AI can add "novel heuristics".
4.) Any query plan, runtime-efficient or not, will produce the same result. AI is not messing with results.
5.) Adding further indices to a DB schema will quickly be a double-edged thing, as index maintenance will also consume serious runtime.
"The most expensive thing that can happen to a government agency is to lose sensitive data to a competing government.
For details, you can ask Karl Dönitz and Isoroku Yamamoto.
Windows must be banned from processing any secret government information, as they are at least 20 years behind the state of the art."
Open Source does not have the bribery infrastructure in place. Unlike the $corporation.
Also, good people avoid the government, as it cannot pay competitive wages. Instead they hire armies of losers.
When push comes to shove, a college-dropout oligarch will be the Effective Surgeon General, as we have seen with COVID.
A dark world full of corruption.
Oligarch bribing newspapers: https://www.berliner-zeitung.de/news/gates-stiftung-unterstuetzt-den-spiegel-mit-weiteren-29-millionen-dollar-li.194183
Oligarch buying goverment:
https://www.infosperber.ch/wirtschaft/konzerne/who-geraet-immer-mehr-in-abhaengigkeit-von-bill-gates-co/
https://www.welt.de/politik/deutschland/plus209247817/Umstrittene-Finanzierung-Das-Gates-Dilemma-der-WHO.html
Of course this is absolutely, never, ever related to his Pharma Investments !
CISA coordinated with Facebook, Google, youtube, Apple and others in order to squelch any anti-Covid messages of ordinary NATO citizens.
https://judiciary.house.gov/media/press-releases/new-report-reveals-cisa-tried-cover-censorship-practices
https://eu.usatoday.com/story/money/2023/10/03/fifth-circuit-cisa-ruling-biden-first-amendment/71051110007/
https://www.theregister.com/2023/10/04/cisa_barred_from_coordinating_with/
It was essentially a CIA-NSA-ARMY operation designed to facilitate illegal censorship. They seconded their operatives into CISA, so they could avoid legal trouble.
This backfired big time, as people simply used TELEGRAM, provided by Mr Durov "out of Dubai" (believe this at your own cost).
Yes, that's true. Free speech provided by Russia because CIA censors the h3ll out of American services.
You can do "fancy" things with waveforms, you can use redundancy in the coding, you can use secret waveform creation(depending on a key) and you can use directional antennas.
All of which helps massively against jammers.
But in the ideal case for the jammer (being very close to the receiver), this does not work. That's true.
I realize I am behind the bullshit cycle.
It's actually a useful concept: https://www.coursera.org/articles/data-lake-vs-data-warehouse?utm_medium=sem&utm_source=gg&utm_campaign=B2C_EMEA__coursera_FTCOF_career-academy_pmax-multiple-audiences-country-multi&campaignid=20858198824&adgroupid=&device=c&keyword=&matchtype=&network=x&devicemodel=&adposition=&creativeid=&hide_mobile_promo&gad_source=1&gclid=EAIaIQobChMIjb-ii-TEhgMVCbRoCR0qyBAxEAAYBCAAEgLQCvD_BwE
Apparently it is a factor of 100 slower than LISP for general purpose computing tasks like Prime Number calculation. SQL was never intended as a general purpose language, so that's OK for me.
In the imagined "SQL/LISP" environment, SQL would be used for "data reading and writing from permanent storage" while LISP would be used for "complex computation".
There would be at least three software engineers who could make good use of this approach ;-)
Isn't PL/SQL an imperative/procedural language running inside the Oracle query executor?
It would be surely interesting to have LISP/SQL in the same setting, though. Lisp is functional.
See this page for language classification:
http://sappeur.di-fg.de/classification.html
First clean up the SSL/TLS abomination. It is designed to be very hard to implement securely. For a long time it was effectively an open front door. There exist much better alternative concepts from Europe.
Then proceed to smoke out C in the kernels, go for microkernels. Have a look at Oberon, its ingeniously compact.
But do you really have the will, the minds and the money ?
The tools for secure comunications are already part of Linux:
+ end to end GNUpg encryption. Never trust a $hitty email server.
+ AppArmor Sandboxing of all apps from Firefox to LibertyOffice to Thunderbird.
+ Dump Microsoft, as they are easily 20 years behind the state of the art. Sandboxing for starters. This is the only language they understand.
+ Never trust the cloud for anything sensitive or more. Rather, run ssh/scp based file servers INSIDE your network.
+ Monitor all traffic at the firewall, maybe AI can detect unusual patterns there.
