Ask HN: Help redesign a Comp Sci major

[mitguy]

I have the task of refocusing the Comp Sci program at a small liberal arts college, with an even smaller Comp Sci department. The trad MIT/CM/CalTech theoretical program just doesn't work for these students. Emphasis should be on employable computing skills that can be delivered almost entirely by 1 prof. What should I teach? What can be cut out? What would you call the major?

[EnderMB]

I'm absolutely shocked that people are okay with this.

One of the biggest problems I've found with recent graduates is that the programs they've come from have been nothing more than trade schools for Java and PHP developers, and most of them lack even the basic underlying knowledge of things like algorithms, paradigms and how to make efficient code. This is not Computer Science, this is basic programming and is no better than what you'd get from the numerous video lectures on the web.

In short, employable skills are best left to employers, and academic subjects are best left to academia.

If it were my choice I'd make Computer Science a purely theoretical course from the outset. If these kids don't have a background in Math give it to them so that they can read through the likes of Introduction to Algorithms and TAOCP with no issue. Don't even let them touch code unless it's to illustrate a theoretical aspect of CS.

Once they've got a solid year of CS behind them and they're all of a sufficient standard in the theoretical aspects of CS then they can be introduced to programming. However, I'd steer clear of the employable languages and force them into Python, Lisp, Haskell, Prolog, R and co to reenforce the theoretical background.

Of course, you need these graduates to be employable, so why not offer a course titled "Internship" during the second year? Let them spend a few days each week working in a real company with real developers that can teach them real skills? I graduated from a modest CS programme and thanks to my numerous internships I was offered interviews left right and centre. I learned more about programming in an eight week internship then I did in an entire year at university, despite most of my second year being about learning to program, and because I had real world experience in employable languages the university didn't cover like C# I walked into a job while others in my class with better grades struggled.

You get the idea. Please stop turning academia into a trade school and teach these kids real CS. If you want them to be employable offer a mandatory course where students have to work in a real business and are forced to pick up programming from people that actually know what they are doing.

[mitguy]

I do like the idea of an internship!

Real-world experience can not be overvalued. That being said, I'm still working with the parameters in my original post. Small. Liberal Arts. College. This isn't a huge program, I've consistently had between 12 and 20 students in the major, of all grade levels and wide differences in skill levels, experience, exposure, and mathematical prowess.

I know it may be hard for others to understand what my student landscape looks like, but after more than a decade, I feel I understand them. But a year of strictly theoretical CS, and I'm looking at 2 to 7 students in the major, and filling out my teaching schedule with writing classes.

Looking at a newly built house, you see that all levels of craftsmanship have been employed. If every contractor is a framer, because it's the fundamentals of building, your interiors look pretty basic, and let's not even mention plumbing made out of lumber!

What I'm trying to say is that I've tried my own version of CS snobbery and highmindedness, and it isn't working. I'm looking to design a computing major that isn't producing only framers.

I appreciate your input, please reread this post and my original post, and try to put yourself in my position, then send me some more suggestions!!

[EnderMB]

Like many students, I studied at a modest university with what I was told was a good CS programme.

It turns out that the reason it's good is because it teaches students how to become the next batch of Java and PHP developers for local businesses. The actual CS content was minimal and bizarrely a lot of classes from the "CS school" were out of bounds to my CS degree, because learning XML was more important than Discrete Math.

I spent half a year studying at a top 5 university and the difference was overwhelming. The facilities weren't much better, but the lecturers were far more helpful and the content was not watered down. Perhaps it was because it was a Masters programme, but the lecturers were happy to help, even when you were hopelessly stuck. Sadly due to money reasons I couldn't stay, but I'll definitely return to get my Masters when I finish at my current job.

It still amazes me that a lot of students don't opt to take internships during the summer. I worked for eight weeks after each academic year and everything I picked up was invaluable. The best part about it was learning what you were good at and what you sucked at. It taught me languages I'd never learn at university and it taught me exactly what is expected of a developer in the real world. Most importantly, I was employable right off the bat. My degree almost didn't matter because the skills I had gained through these groups of eight weeks were what got me my job.

I appreciate that you are constrained in what you can do, but I do feel that a practical approach to CS and general trade Computing is less than perfect in an academic setting. If possible, would you be able to highlight the exact problems you faced when trying the typical theoretical route?

[codegeek]

"employable computing skills "

I think this is a very broad statement even though not trying to undermine your question. The problem however with this statement is that it varies. Some people have mentioned HTML/JS etc. which are surely good but not all tech. related jobs need that.

Some things that I strongly feel should be part of every CS curriculum especially in the context of your question i.e. employability are following:

- Understanding real world projects and the lifecycle. This should include practical experience delivering software and manging the entire phase. Some universities already have a "software engineering" class for this.

- Related to #1, teach them the value of teamwork. No real world project can be done alone and the bigger the employer, more likely you will deal with cross matrixed teams (yea teach them what cross matrixed means :)

- Teach them about distributed computing and how software is deployed/managed in a large scale environment.

- Teach them that no one cares if you build cool stuff. People care about what problem it solves. It could mean anything from keeping your company/manager happy to solving real world client problems.

- lastly, tell them that whatever you learn now, you will learn a heck of a lot more when you are actually out in the real world and that will make you realize that you did not know anything. This does not however mean that the learning is useless but it is just the tip of an iceberg.

[mitguy]

Thanks! Good points, CG.

I know the "Give a man a fish" adage, but trying to teach my students to be Global Anglers hasn't worked well for a decade (given our students, numbers, faculty, and facilities). I'm not shooting for trade school either, just something that's very mindful of your iceberg analogy.

Tip: - Come, learn some skills, peek at the landscape.

Iceberg: - Go get a job(s), learn the rest of what you need to build the solutions that keep your company/manager happy.

[jfaucett]

I think unless you're a computer scientist or engineer at google, fb, etc. most of the theory you run into at the university level you're never gonna use. Having said that some "theory" is important for me as a developer, but really it just boils down to 2 subjects Discreet Maths and algorithms, so I would try to emphasize those two areas in the program.

As far as employability goes, I think having a strong understanding of networking (TCP/HTTP) and Data/Databases, is really the core ( ie Data & Application layers) of most development these days.

Also I know its kind of experimental but if possible I would actually teach go as a programming language instead of the usual c/c++/java (for the systems level / networking stuff). Go in tandem with javascript (both server and browser) for web applications, allows you to deal with just about every programming concept there is at an abstracter level and both languages are not overly difficult to grasp.

So to put these thoughts together. Maybe you could center the program around networked software development? This lets you teach things like webkit, browsers, js, networking, servers, and databases while all staying in one (though very broad) medium. I think this could be possible for 1 prof. So for instance, in "interface design" you're in the browser and delve into webkit and user interaction, or for "computer graphics" you can go into "webgl". For "databases" you're still dealing with server / client architectures for which you can examine db structures different types of web applications.

I'd love to have had an option to do a pure web engineering compsci program. Maybe some of my ideas will help, good luck!

[rhsanborn]

You may never use most of the things you learned. But they almost certainly had an impact on how you thought and developed. I'm surprised how many times silly theory of computation things come up and maybe don't force me to change my solution, but get me to think of it at least a little differently.

[jfaucett]

Yes, maybe I should have mentioned proofs/theory of computation seperately I kind of just included them under the general algorithms umbrealla term :)

[rhsanborn]

I agree with several of the posts below, but I have to take it a step further. I'm not sure you can or should call it a major in the traditional sense. It cheapens the idea of a degree if we teach people trade school curriculum and give them degrees. I've encountered a lot of these people, and it's the reason a lot of people from private institutions (Baker, Davenport, Phoenix, etc) get passed over for jobs.

If people really want a degree in MIS or CS, then the curriculum should reflect that. If they can't handle the material, then I think the college should look at the possibility of offering certificate programs in specific areas.

A student who understands theory can generally go out into the world and adapt. A student who understands how to make a web page using specific languages and specific steps and can't go much beyond that hasn't learned the things necessary to be given an associates or bachelors label.

Question: Are the students capable and don't have the prerequisite knowledge? Maybe the college needs to consider an entry exam and remedial classes on basic computer use, theory, and logic? If they aren't actually capable, then I stand by what I said above. I think the college should consider certificate programs like, "Web Programming with .Net and Javascript".

[iamshep]

I run a 13 person IT shop at a Big-10 university. If I were a single person shop, what would be different? I wouldn't write code. I would be all about integrating off the shelf software into my systems. I would very likely be stitching University databases and third party apps together. I would be a call center and a support shop. I would be doing a lot of reading logs, Googling errors, and generally trouble shooting. I would be more than functionally literate on MACs, PCs, Unix, and MF machines. I'd know a reasonable amount about networking, and even more about SQL.

If you were going to roll this into a minor, I'd call it something like Applied Information Technology Management (or some mix of such buzz words). (Note: James K. badgered me into posting)

[tjr]

Some sort of class on embedded programming. Lots of software development jobs are for embedded programming, which is in many ways a world apart from web applications. Things like avionics software, automotive software, medical device software...

Recently the JPL C coding standards were posted around the web, as people showed a bit of interest in the software running the recent Mars robot. Discussing that sort of thing could be interesting course material.

I don't know of any curriculum off hand, but maybe make it some sort of project-based course where students design and build real-time applications on some sort of real-time Linux system?

[brycecolquitt]

If these are the type of students I'm thinking of, they really don't want to become computer scientists, or software engineers, even. They want to learn how to code so they can build stuff, namely web apps. I'm assuming that since they don't want/can't handle a theoretical program, yet still think they want to learn Comp Sci, they fit this description.

So I'd teach towards that:

1) an understanding of the technology stack for a web app 2) Javascript 3) Ruby or Python w/ Rails or Django

[mitguy]

Yes, I think web apps is a significant part of it. Looking at projections of software sales, web apps and web app support are on the horizon for almost everyone.

[probinso]

It sounds like you want to seperate software engeneering and computer science. Computer science is a different course material.

[mitguy]

Agreed!

But I'm looking for a hybrid. I don't want to abandon my theoretical roots, but I'm missing more than I'm hitting, with my students as of late.

Plus, you can have Computer Science without Software Engineering, but the reverse isn't true.

[ghenne]

Good idea! Some things to teach...

HTML JavaScript programming, Practical SQL/SQLite usage, Project Management, Visual Studio .NET

[mitguy]

I think .NET seems a bit heavy for what I've got in mind. I also probably don't understand .NET as well as I should either.

Can I get some more clarification why .NET should (or shouldn't) be included in a modern, practical curriculum?

[tjr]

Teach this class: http://philip.greenspun.com/seia/

[mitguy]

Yes, this is definitely a course type on the radar!

It echoes some other responses, and also some of my thoughts.

MIT Press makes it that much cooler!

[ghenne]

Some things not to teach:

Don't do: Numerical analysis, Compiler design, OS programming, Real time coding

[mitguy]

As much fun as Org/Arch is, I'm thinking we can leave most of that and ASM behind, as well.

Not totally behind, but committing a semester to them as a course is what hasn't been working.