Comments by "Vitaly L" (@vitalyl1327) on "Continuous Delivery" channel.

  1.  @dacianbonta2840  another "let's just glue the libraries together" self-styled developer detected. How hilatious that your lot really believe you have a right to have opinions.
    1
  2.  @dacianbonta2840  you're not very coherent, and you clearly have reading comprehension issues. Looks like you're beyond any hope, no amount of training will form you into even a barely useful engineer.
    1
  3.  @h4ck3rd4wg  two words: message passing.
    1
  4.  @secretchefcollective444  the problems are not hard - go to shadertoy and see how easy all the graphics algorithms are when done right. Photoshop is a huge pile of legacy code upon legacy code, awfully overengineered, just like 99% of the other code out there. Most programmers have no idea how to program.
    1
  5.  @secretchefcollective444  examples? Think of, say, CADs. Geometry of the B-Rep is insanely complex. Finding an intersection between two NURBS is computationally intensive and mind-bending. But why? Why people keep using b-rep when all the solutions turn embarrassingly trivial in an implicit representation (such as SDF)? Same goes for 2D image processing - just turning your data into a right form eliminates all the complexity from the problem.
    1
  6.  @secretchefcollective444  it's perfectly possible to derive splines from SDFs (for slicers, for any forms of CNC). Much easier than from b-rep. More so, SDF is ideal for modelling the tool path for CNC. And yes, mesh-based systems are overengineered too.
    1
  7. Not all languages. Some are vastly more awful than the others. Good languages tend to be designed by people who know what they're doing, who have a deep knowledge in PLT and in compiler construction. Really bad languages, like Javascript, PHP and Python, tend to be designed by eager amateurs. Somewhat less crappy languages are designed by amateurs but later fixed (or at least specified) by professionals - case in point here is Java, with spec written by Guy Steele Jr.
    1
  8.  @chrisjsewell  I'd say both Python and JS are exceptionally awful scripting languages and must be avoided. There are far better options.
    1
  9.  @chrisjsewell  various Scheme implementations, for example - including the official GNU scripting engine, Guile, and more performant alternatives like Racket. LuaJIT is also very good. Raw Lua is very small footprint and still very capable. Various Forth implementations can be extremely small footprint and yet nearly infinitely capable due to the meta nature of the language. Even Tcl have its place, and often is a right tool for embedded scripting (in this one I might be biased, as nearly all EDA tools are scripted with Tcl).
    1
  10. There is only one meaning to "M.A.D." - mutually assured destruction. People must stop trying to fit anything else there.
    1
  11.  @ApprendreSansNecessite  language with separate statement and expressions cannot be called functional language, full stop.
    1
  12.  @ApprendreSansNecessite  It's either lambda calculus (which only have expressions), or statements. Not both. ADTs have nothing to do with FP (besides being mostly explored by FP-first languages). Recursion in JS is useless for the lack of tail call optimisation guarantee. But first and foremost, FP = lambda calculus, and lambda calculus is expression-based. Another mandatory thing for FP is strict static scoping. Everything else is optional, but expression-based language, unlimited tail recursion and strict static scoping are mandatory.
    1
  13.  @ApprendreSansNecessite  there is an important property in the pure lambda calculus that is missing from eager languages - it assumes lazy evaluation (to be practical). If your language is eager order, you cannot have a pure lambda, you need at the very least a built in 'if'', along with 'true" and 'false' in some form. So you already cannot get to understand the most essential properties of lambda calculus if you're doing it in an unsuitable (too big and complex) language.
    1
  14.  @ApprendreSansNecessite  later I will produce an example when eager order makes it impossible to evaluate a term (hard to do so typing on a mobile). Sure, learning FP with whatever tools you have is better than not learning at all. Just keep in mind that results of such learning may be quite incomplete. JS is very far from an "ideal" language for anything, not just learning FP, and the only virtue it have is (and you seem to agree) is that a newbie is likely to already know some bits of JS due to its popularity and accessibility. It is a new Basic in this regard. Which is rather sad.
    1
  15. Take a look at Mathematica, and you'll never think of Python as more than a toy.
    1
  16.  @twodyport8080  what planet you are from? Because on this planet experience is the opposite - impoasible to find competent developers, recruiters and head-hunters flocking around every single developer they think they can sell. You're obviously doing something very wrong if this is not your experience.
    1
  17. it's more about the real quality of the majority of developers and architects than about some perceived capabilities of this so-called "AI" (in fact, just statistical predictive model). If a majority of code monkeys are worse than a dumb predictive model - well, they should not have been code monkeying to start with, and when they're all gone - good riddance.
    1
  18.  @ApprendreSansNecessite  it's a really bad term. Compilation does not define what's the level of the source and target language. A lot of modern compilers consists of dozens of tiny compilers that translate a language into a nearly equivalent language, with some minor changes. Adding some babble like "transpiler" to the mix just confuse people and obscures the meaning of compilation.
    1
  19.  @ApprendreSansNecessite  the word "transpiler" is from colloquial use indeed, you'll never meet it in any PLT papers. So it's important to remind people once in a while that the term is unscientific and is better avoided altogether for adding no real value.
    1
  20.  @MartyrCry  this may only apply to the very junior roles. I rarely see stack even being mentioned in the role descriptions (and I certainly never mention it myself when I write such descriptions), for it's an irrelevant tiny detail.
    1
  21.  @catocall7323  this is the consequence of overabundance of very low quality junior developers from bootcamps and such, they are not generalists and will never be able to turn generalist, they are trained monkeys tuned to a single stack and a very narrow range of problems. Industry adapted to this situation, making it very hard to the few worthy ones to get in and to get noticed at all in the huge pool of the bootcampers. This is an awful sitiation indeed.
    1
  22. @ wrong.
    1
  23.  @fulconandroadcone9488  Like, a 6 years-long onboarding process for juniors to go through all the CS curriculum they've missed?
    1
  24.  @arcomarco7131  there is. Or you know where I can find competent engineers? I am all ears.
    1
  25.  @guamie  agreed, experience correlates with quality but does not guarantee it.
    1
  26.  @kevinmcfarlane2752  there absolutely is a shortage. And it may be caused by how much the pool is diluted by incompetent ones, making it very hard to find the good developers among the masses of mediocrity. Works the other way around, good developers fail to find jobs for they're drowned in the said masses.
    1
  27.  @yugiohfanatic1964  if the same work can be done for 1/10th of the price, it must be done for 1/10th of the price, obviously. The problems start when it turns out that it won't be the same work, not even close.
    1
  28.  @dacianbonta2840  look, no company can train a developer. Never. A good engineer is a product of at least 6 years in a higher education, with a few internships and an active community projects participation. The problem with junions these days is that pool is so diluted with hopelessly bad ones - all the bootcamp graduates, people from second tier universities, etc., that it is impossible to find the good juniors I described above. Most companies simply gave up and don't even try any more. There is absolutely no way companies could step in and replace the function of the proper universities
    1
  29.  @dacianbonta2840  it takes 6 years of a university to prepare an engineer. You now want companies to do it in a few months instead? Really?
    1
  30. The tech world does not need (and really never needed) juniors churned out by all those bootcamps and sub par universities. They were always useless. Smart fresh people from top universities, with solid STEM degrees, will alway be in demand.
    1
  31.  @gauravtejpal8901  found a bootcamp "graduate"
    1
  32.  @gauravtejpal8901  found a bootcamp "graduate".
    1
  33. oh, really? Any developer must master dozens of languages. Any capable developer must continuously create new small languages, because Language-Oriented Programming is by far the most powerful abstraction tool. A developer with only one language under their belt, especially such a crippled and ill-designed language as JS, is not a real developer. Want to be "full stack"? Master the entire stack underneath your level of abstraction. Starting with analog and digital electronics, HDLs, all the necessary elements of computer architecture, all the PLT-related mathematics, PLT itself, a deep understanding of compilation and interpretation methods available, a deep understanding of how operating systems work (and when and where to use them or ditch them altogether). Understand networking, from PHY levels to high level protocols, and know where you should be on this stack of abstractions depending on your goals. There's a lot of critically important knowledge there. Javascript, browsers and all that meh are not among even the tangentially impactful pieces of knowledge.
    1
  34. To summarise - the language you must be a native speaker of should be the language of mathematics. All the programming languages are just special cases of it, once you have a solid grasp of the real language of the universe.
    1
  35.  @ApprendreSansNecessite  the downsides of compiling to JS and alike is that the runtime itself is rather crappy (no integers, really? and a very intrusive GC), and, more importantly, if JS is your common denominator language, it limits what your static analysis can do (pretty much nothing for such an extreme dynamic language). The only reason to ever do such a thing is if your code must run in a browser. No point in running JS anywhere else.
    1
  36.  @ApprendreSansNecessite  I'm yet to see a single compelling argument for using the same language for different components of a system, be it "client side" and "server side" in a context of web development, or anything else. And the argument "our engineers are just not good enough and can only use one language" is really a bad argument. And yes, I'm aware that there are web developers who style themselves arrogantly as "full stack" developers. They need to be constantly reminded how thin and tiny their coverage of the tech stack is, and how much more they could have achieved if they were actually full-stack.
    1
  37.  @ApprendreSansNecessite  in a multi-language environment you're limited to an underlying target language semantics for your analysis. And if its semantics is designed to hamper any analysis attempts, you cannot prove anything about your system as a whole. Typescript does not help, as it's not a compilation target for any of the higher level languages you're talking about. Also, even Typescript does not remove the dynamic nature of the underlying language, so even a heavily annotated typescript code cannot really be reasoned about.
    1
  38.  @ApprendreSansNecessite  for me, it sounds like a weird design, when similar data processing happens in different parts of the system. For data model - yes, it can be the same while flowing through the system, but it does not mandate the same language that handles the data in any way. Validation is normally just a part of the data model. And things like data model, validation, etc., should be defined in a higher level (ideally, declarative) language anyway, and then translated into whatever language whatever part of the system is using. You're likely familiar with rudimentary forms of this approach employed in IDL and similar protocol / data model description languages. There's often dozens of languages involved, with a single data model in between them. I admit I stay away from anything related to web, so may be unaware of some of the rationalisation behind common design choices in that world. From far away, the whole web stack looks like it's massively overengineered and badly designed on all levels though.
    1
  39.  @ApprendreSansNecessite  I believe this sort of behaviour still falls under data model specification and must be declarative. When it is some ad hoc low level code manually screwed together, it is unavoidably a huge source of errors and worse. I never cared at all about anything web. But I did a lot of work with distributed systems shuffling huge variety of data around, including very sensitive data in financial applications (that also included high frequency trading). Every single time, some form of IDL was the only reasonable solution, and sharing validation / processing code between system components was never a meaningful design choice.
    1
  40.  @davidjulitz7446  you do not need to prove the whole system, just the distributed communication part. This is what TLA+ is for. And this way it'd quicly becone apparent that there exist a number of conditions where network lag or breakdown compromise transactions integrity. Every communicating system must go through such a proof. And we do it for quite complicated, heavily distributed designs.
    1
  41.  @tonymerrett  I've been building fault-tolerant systems for over 30 years.
    1
  42. Did it occur to you that you can be the problem here? It is very hard to find developers in the EU, the shortage is very obvious.
    1
  43.  @tacorevenge87  France, Spain, Germany. Outside of EU the market is very lively in Switzerland and the UK.
    1
  44. This is why eDSLs are better than pretty much anything else. The most readable code is the code free from leaky abstractions, and you can do it only with language-oriented programming.
    1
  45.  @langhamp8912  yep, that's my point exactly. Though I doubt you can teach the scientific method properly without using any of the hard sciences as a guiding example.
    1
  46.  @nerdobject5351  not sure about zero wear and tear - sitting is the new smoking, after all. And RSI is real.
    1
  47. It can't even do guaranteed TCO. How can you FP with it?
    1
  48.  @skyhappy  first, let me reassure you, that you have no faintest understanding of the real world. It's pathetic, frankly, how code monkeys always run around rambling about the "real world", why they don't even possess a mental capacity to even start understanding it. Secondly, no, computer science is the most important science out there, binding all the other sciences together. It's the science of what computation is, and our universe is built on the notion of computation on many levels. You won't understand it, of course, I'm saying it here not for your benefit. I'm a nuclear physicist who had to turn to computer science, because there were no answers to my questions anywhere else. And only computer science could finally make all the pieces of the puzzle click in. And computer scientists are supposed to explore this branch of knowledge. Not write some pathetic crappy code, like your kind does. And I'm so glad I'll never have to work with primitive nobodies with overblown ego like you. That's another great feature of the current market, it's very picky, and it works exactly as intended, leaving your kind out of anything meaningful.
    1
  49. One have grey hair, and the other does not need little blue pills to perform.
    1
  50.  @edgeeffect  some languages were designed by PLTists, e.g., Scheme.
    1