[lsc]

>I simply do not understand this anti-college anti-credentialling sentiment

Speaking as an anti-college, anti-credentialist person, well, a lot of it, for me? is that I didn't go to college.

I mean, everyone likes to pretend to be altruistic, but in the end? we all see the world from our own perspective.

And really? most people seem to think that you go to school, you get a cs degree and you will be pretty good. And that's simply not true. I've hired people with CS degrees only to have to fire them, because it turns out, they didn't actually learn anything.

I mean, I've also worked with people who really did learn incredible things in school. Things, as you said, that I certainly have been unable to teach myself.

I also think a lot of this anti-college stuff is people who went in to debt for life, having been told they'd get useful job training and, you know "find yourself" (you know, the class thing. No matter how much money I earn? I'll never be middle class; in the eyes of most people, i'll never be a 'professional' a 'real person' unless I go get a degree. I'm the guy who fixes the pipes. Which, eh, I am mostly okay with at this point.) - but anyhow, yeah, these kids graduate and find that the market value of their degree is, well, pretty close to zero. (and yes, a lot of these kids got degrees in philosophy or fine arts... And yes, we laugh at them for expecting to turn those degrees into money. But they were 17, goddammit, when they made that decision. What kind of asshole expects a 17 year old to make good decisions, especially when they receive bad advice from every trusted figure they talk with? Then we go and say "unlike every other bad loan you will take in your life, the person giving you the loan takes no responsibility. You have to pay this one off. No bankruptcy.")

This idea that "you can do anything you set your mind to" also permeates education, and it's incredibly destructive; it's how we end up with those people with CS degrees who can't outcode me.

[necubi]

You don't learn to code in college, or at least not in class. Good coders almost universally learn outside the classroom. This is true for people who go to college and those that forgo it.

So if you don't go to college to learn to code, why go at all? You go to learn the fundamentals of computer science and math which underlie everything that we do.

It's certainly possible to obtain this knowledge on your own with serious self-study. But it's a hard route (albeit one getting easier thanks to Coursera et al.) and given two candidates with equal coding ability, the one who went to college is far more likely to have a strong grip on theory.

Is theory actually useful? I would claim that it is, even for programmers doing CRUD apps and the like. Knowing the fundamental abstractions of computer science and the ways people have applied them in the past saves an enormous amount of needless reinvention. As a concrete example, parsing is a very well-studied problem in academia. But somebody without that background trying to write a parser will struggle far more than another familiar with CFGs or PEGs.

Being able to program is just scratching the surface of computer science. It's necessary but not sufficient for being a great programmer.

[lsc]

>It's certainly possible to obtain this knowledge on your own with serious self-study. But it's a hard route (albeit one getting easier thanks to Coursera et al.)

Now, what little math and theory I have learned, well, really, what anything I've learned... I've found that it's dramatically easier to learn from books than from a lecture. First, people talking? really, really slow. I ain't no speed-reader, but I can manage 400 words a minute, give or take, without skimming; of course, I can go much faster when I'm just skimming over the stuff I already know.

My usual strategy is to buy several books at the same level on the same subject; Preferably by very different authors. (I want at least one of the books to be a classic by the person who came up with the idea.) the idea being that each one is going to explain it in a slightly different way, and while I'm only going to retain a small portion of each book, well, books read fast, they don't cost much, and if I take away 10% of each book? I'm doing pretty okay.

(I mean, that said, my math and my theory is still pretty weak, compared to many of my educated peers who work at the same payscale, So I'm not saying my strategy is good or anything... it's just that I've never gotten much at all out of lectures. Payscale is also interesting... it seems to scale more with negotiation aggression than anything else.)

>Is theory actually useful? I would claim that it is, even for programmers doing CRUD apps and the like. Knowing the fundamental abstractions of computer science and the ways people have applied them in the past saves an enormous amount of needless reinvention. As a concrete example, parsing is a very well-studied problem in academia. But somebody without that background trying to write a parser will struggle far more than another familiar with CFGs or PEGs.

Why the fuck would you write your own parser for a CRUD app?

Seriously, that's whats wrong with the world.

I'm a SysAdmin. The way I see the industry? you hire a bunch of php monkeys to slap something together. customers don't like it. You pivot, come up with some other business idea, hire php monkeys to slap something together (usually using the same monkeys and the same code) you repeat until the market likes the output.

By this time? It's a giant hairball. It's disgusting. But you are getting users, and it's making money, so you've gotta scale.

Now, you could pay a bunch of 'real programmers' to write you something good, but that can take months or years even with competent management.

So what do you do? you hire someone like me. I show up and put a caching http proxy here, I put some indexes on the database server (and put it on a beefy server, and maybe even setup a read-only replica for the database) etc, etc.... i slap on duct tape until you get around to having a competent person re-write the whole mess. That's the "computer janitor" role.

So yeah. that's why I hate CRUD programmers trying to re-invent the fucking wheel. There are plenty of parsers out there. Use them.

This, this is why sysadmins hate NoSQL. seriously, in a few days I can make a crashy access db 1000x faster and more stable fairly easily by ODBCing the data out to a reasonable database. relational databases are incredibly easy to tune and improve, and I don't need to understand your giant PHP hairball that is full of global variables. USE A RDBMS. then you can do your job, and then if the thing takes off and you need to scale, I can come in and do mine.

[necubi]

That's sort of my point. People without a computer science education often lack the awareness of the gaps in their knowledge. So they don't set out to write a "parser," because they don't know that word.

They're faced with the problem of parsing some custom text format. Maybe they'll try some terrible regex hack (see: http://stackoverflow.com/a/1732454). Maybe they'll stumble onto recursive descent parsing. But if you've taken a class on compilers (or automata theory) you'll know when regular expressions are not up to a parsing task, and you'll know about parser generators and how to write a grammar for one.

How can these CRUD programmers avoid re-inventing the wheel if they've never been exposed to it?

[pkolaczk]

"relational databases are incredibly easy to tune and improve"... only as long as they are on a single beefy server. When it is not enough, and very often it is not enough, things start to become very hairy up to the point where you have to rewrite your application using a proper NoSQL solution and with some people who understand databases a little better than "See ma, I put SQL here and it automagically evaluates..."

[lsc]

servers up to 512GiB ram, these days, are pretty cheap. that is /big/

The thing is? last I looked, none of the NoSQL databases handle multi-master replication, either. In all cases, if a server goes away, you have issues. With SQL, you promote your most up to date read-only replica to read-write master, re-point the other slaves and run. With modern SQL databases, you can even control how out of sync you allow your read-only replicas to be (to the point of making it so the transaction isn't complete until it's been written not only to the master, but also to the slave... of course, that costs time. It's a tradeoff you can tune, depending on how much your data matters.)

Now, if you just need more write-performance than a single beefy server with 512G ram and a bunch of SSDs can bring (which is a lot) then you shard. At that point, you are pretty much where most NoSQL systems are, except that most sql engines come with a bunch of neat tools for finding slow queries (and for figuring out how to make those queries fast without changing the php) that the nosql stuff seems to lack.

[icelancer]

>So if you don't go to college to learn to code, why go at all? You go to learn the fundamentals of computer science and math which underlie everything that we do.

This is why my company hired Math/Physics graduates instead of CS graduates - they had far better command of advanced mathematics and came much cheaper than CS graduates (who were generally worse applicants anyway).