It's not as bad as you'd think in total volume, it's just you get a lot of bursty concurrent traffic. Something as simple as a hundred requests in roughly the same second crash many blogs sadly.
The most important lesson I've learned is simplify. Simplify everything. Simplify your problem. Simpilfy your approach. Simplify your architecture. Simpilfy your data model. Simplify your build process. Simplify your release process.
Complicated things break. Simple things also break, but a lot less often.
The problem is "simplify" means different things to different people. For some it's write everything in vanilla JS. For others it's install libraries for everything, write the least code.
"It doesn’t matter if you don’t know algorithms…
But it doesn’t hurt to learn about it (because it might help you in the future).
This is important. Usually you don’t use algorithm in programming a system but if it’s something that will scale fast, your understanding of algorithm might be useful."
Sigh. I'm really tired of this misunderstanding.
Literally everything you type in to an editor is an algorithm. "Learning algorithms" isn't about memorizing known facts, it's about learning how to think about programming.
Those that don't study algorithms are doomed to reinvent them, often poorly.
Totally agreed. A knowledge of "conventional" algorithms influences how you write code. Sometimes it manifests as knowing which algorithm to choose, sometimes it manifests as intuition into the performance (or lack thereof) of something you're writing.
It's not required by any means, but don't turn down cheap knowledge on how to gain an edge.
Some basic algorithmic knowledge is definitely essential. I remember the first time that I found a need for sorting. I designed my own algorithm, and I learned years later that I'd basically recreated bubble sort. It worked fine in my little QBasic program, but would've made anything serious fall flat on its face.
I remember a recruiter that tried to convince me some job would be interesting for me: "It's really complicated work. We use algorithms." after which she dropped a dramatic pause to let me recover from that shock.
Don't write/use abstractions you won't understand later. Often you start with 2-3 copies, try to merge them and end up using it at a place were it doesn't fit 100% anymore.
Keep nesting low. Often it's faster to scroll over one big okay-ish structured function than to find all the clean tiny things around the the project. Same goes with Objects/Classes/Components etc.
Stop using abbreviations and try to write code more readable and not as short as possible. IDEs help with autocompletion and code gets read more than written anyway.
This so much. One of the bigger differences between academic and professional coding is software system longevity.
Premature refactoring is incredibly harmful to readability.
1. You have no idea what the future will require. You will always know more about the requirements in the future.
2. You have no idea how the current use and system will need to evolve. You will always learn more about this in the future.
3. You have no idea how long the current system will be used. Could be 20 years; could be a week (if it's an executive's pet project).
4. Time to MVP > perfect code in almost all instances.
Writing code with clean, documented potential refactoring points, but stopping short of actually pulling all those things apart has served me far better than the alternative.
If you have one function that may not be applicable to a future use case, and you refactor it into two, now you have two functions that may not be applicable to a future use case.
Develop a feel for what's "reasonable" in an initial write, and don't code yourself into an inflexible corner, but defer actual refactoring until it's necessary. Because then you'll have the benefit of knowing far more about the new needs and how the system is used.
> Premature refactoring is incredibly harmful to readability.
Is that a thing? If it is, I've never run into it. I always run into the opposite - never refactoring and everything turning into a god object. It never occurred to me that premature factoring would be bad because I can't recall ever running into it!
I've seen plenty of this. The most common offenders are junior to mid-level developers who, in the pursuit of attaining the trappings of sophistication, cargo cult various patterns without understanding the original rationale behind those patterns, nor when they should be applied. What they fail to realize is that needless indirection is not the same as useful abstraction.
Think what you will of Zed Shaw, but he actually has a pretty good piece on how indirection and abstraction get mixed up:
This is one of the funniest headlines I've read considering the site seemingly crashed. I'm hoping the punchline was actually "Error establishing a database connection" but I doubt it.
Siege [0], is my go-to. Easy to add to CI, because its on most platforms, and has decent output configs. You can vary headers and pages to hit, and how long connections get held open, or just reject them. And of course, it has great concurrency.
It can bring down most servers that aren't configured to prevent DOS attacks, like naive nginx or Apache configs. Usually somewhere between 5-10mins.
a long long long time ago (~2009) i used jmeter. it spawns aws instances and pegs your website in ways of your choosing. by now I am sure they got something even more advanced
When I was on the front page of HN, IIRC it meant about 5k visits in an hour. How can such a low number kill a webpage... Or do other on the front page experience much higher traffic?
- Time don't make you [much] smarter in the "brute force" sense, however it allows you doing less errors, avoiding unnecessary stuff, and in general making you more productive. Also, it allows you to enable others to be more productive, e.g. giving some hint to a coworker so he/she can simplify his work.
- Focus is all: understand the problem, do a good planning, and stick to the plan.
- For critical projects trust people you really know (their capabilities, potential, character, etc.). And DON'T trust people you don't know, in the sense of: tracking, ensuring people know what they are doing and that they're capable, etc.
- The wrong people can ruin any project, no matter how smart you think you are, so: do risk analysis, always.
Everyone is so smug here, wow. If you have a little web presence with 5 visits per day, of which 3 are bots, you don't cache, you don't use AWS or any of the big services, you just spin up a tiny Linux machine with a LAMP stack and that's it.
Is there a good (preferably free) service to load test ones own website? I imagine that something like this is quite hard to test for a beginner and I admit I have never tried to do it on my site either.