[billybob]

If I had to explain the issue at stake here, I'd do it this way.

Imagine you go into a fast-food restaurant. "I'd like a burger," you say to the cashier. You get a burger. The next day, you go to a different restaurant. "I'd like a taco," you say. They bring you a taco.

Just then, a lawyer bursts in. "I'm sorry, but my client, the burger restaurant, has a copyright on that. No other restaurant is allowed to accept orders in the form of 'I'd like a(n) X'".

Would that be crazy?

APIs are the computer equivalent. If you go to a blog and request a page like 'someblog.com/posts', then you go to a movie theater's site and request a page like 'moviesite.com/movies', you wouldn't think of those two actions as having anything in common. Sure, both sites use a url like '/items' to serve up that kind of item. Why wouldn't they?

But if the courts rule that APIs can be copyrighted, the movie site might either have to license the right to have URLs like '/movies', or do something else.

What else? Whatever they can think of - and think of it first. Because the race will be on to copyright every imaginable scheme. '/show/me/movies' and '/movies=all' and '/3932939' will soon be taken. Even if they can come up with a new convention, they'll likely be in court for the right to use it.

Does that sound good for consumers - ostensibly the ones whom intellectual property laws should benefit? Does it sound good for new businesses who don't have legal departments?

Or would it be crazy?

[sunir]

Except that isn't a good analogy. It's much more like copyrighting the Burger King menu, which is the interface to ordering services from Burger King.

By the way, you can copyright restaurant menus.

[ajross]

By the way: you can only copyright the expression of the menu. You cannot copyright the food. If someone wants to offer a straight up clone of the Whopper(tm), there's nothing Burger King can do (though trademark law would of course prevent them from calling it a "Whopper").

[adnam]

Entirely wrong analogy. The order counter is the interface, and the menu is the documentation.

[sunir]

Sure.

Is the source code the software or the compiled binary? What is copyrightable? One or both?

Is an API the source code, the compiled binary, or the running process of the software? What is copyrightable?

Computer software is considered copyrightable. Source code and binaries are both copyrightable.

Even if you didn't copy the source code, if your software is "substantially similar" to another copyrighted software, you could violate their copyrights. The argument is whether or not an API is a substantial part of some computer software.

In this case, it's the Java libraries. Since the Java libraries only value is the Java APIs, I believe it would be hard to argue that copying the Java APIs did not violate Sun's (and now Oracle's) copyright.

That's why I think that it's very frightening that Google has to rely on this argument to defend itself rather than, say, the statue of limitations on copyright or, better, an explicit license grant from Sun.

[chc]

> Is the source code the software or the compiled binary? What is copyrightable? One or both?

The source code is the creative work, and the compiled binary is a mechanical transformation of the work, and thus both are covered by the same copyright. I don't see how this is mysterious.

> Is an API the source code, the compiled binary, or the running process of the software?

It is none of those. The I in API stands for interface — that is, it's an abstract description of the things a library provides and how you interact with them. The source code, the compiled binary and the running process are concrete implementations, not abstract interfaces.

> Since the Java libraries only value is the Java APIs

What on earth? Are you really saying the only value of the Java standard library is the names and type signatures of the packages and methods, and not the actual functionality they provide? The library could actually not implement any of the methods and it would be just as valuable?

I have to assume you are just very confused about what the word "API" means.

[billybob]

It depends on what's considered an API, and that would be a question settled by courts. If a judge decides that someone owns RESTful APIs, every Rails app on the planet will infringe that.

Even if you're right, and the copyright applied only to a specific list of API endpoints and actions, it would still be a nightmare for startups. You couldn't release anything without legal vetting.

[sunir]

You can't copyright RESTful APIs as a class, just like you can't copyright leather-bound books as a class. You can only copyright a specific expression of an idea or collection of ideas. The novel printed inside the leatherbound book is copyrightable, and I would argue strongly that Basecamp's API is also copyrightable and should be copyrightable.

It was possible at one point to patent RESTful APIs, but no longer.

It's not clear to me how declaring APIs copyrightable impacts a startup in the slightest, except of course that their own APIs are copyrightable and they will be prevented from parasitically cloning competitors' APIs.

P.S. There is in fact an company who has cloned Basecamp's API so they could instantly work with all the Basecamp Add-ons and more easily convert Basecamp customers to their own.

[smokeyj]

> Except that isn't a good analogy.

Why is that? If an API can be copyrighted why not a protocol?

[sunir]

Protocols are processes and need to be patented. An API is literally a description and therefore should be copyrightable.

[edit: iphone typo]

[IsTom]

There are declarative ways of describing protocols.

[adnam]

Protocols are processes and apis are descriptions therefore what? You do seem terribly confused.

[sunir]

Fixed the typo. APIs should be copyrightable.

[excuse-me]

A protocol is a set of questions and answers - so is just as copyright-able as an API, or a play.

Shall I compare thee to a JPEG header, thou are more fair and ...

[sunir]

I'm sorry, that's not correct. You can copyright a play because it has a script; the public performance is a specific (copy)right granted on the script. You can also separately copyright a recording of a play.

The main thing is there has to be a specific thing that is copyable. Or more accurately, (U.S.) copyright protects against copying an original work of authorship fixed in tangible medium of expression.

A protocol is not a set of questions and answers, but a mechanism to generate responses. It's literally a machine. Its output changes with every input. Therefore, it's patentable, but is it copyrightable?

You can copyright the software underlying it. You can copyright a transcript of a session of the protocol. You can copyright a technical specification of the protocol.

But because a protocol fails several qualities of "an original work of authorship fixed in a tangible medium of expression", it is not copyrightable.

An interesting and overlapping area are generative works like board games. Take for instance Magic the Gathering or Settlers of Catan. Are they copyrightable?

The answer is more interesting than you might think.

A game is not copyrightable.

* http://www.copyright.gov/fls/fl108.html * http://boardgamegeek.com/blogpost/2700/intellectual-property...

However, the rules, board game, cards, etc. are copyrightable. Does that mean I could make a million billion dollars selling Pioneers of Penzance by directly knocking off Settlers of Catan? No! If the new rules, board, cards are substantially similar to Settlers of Catan, I'd be in trouble.

[excuse-me]

so a print command that takes a format argument and outputs a string is copyright able - a send command that takes an argument and transfers a string isn't ?

[noonespecial]

The problem is that there really isn't a good analogy for this. Its not the menu, or the counter. Its more like going up the the counter, reading the menu, placing the order, going into the restaurant, there being a restaurant in the first place, etc. It regresses away from you as fast as you can analogize until you end up at machine code clicked in with binary switches and some nitwit telling you he "owns" the op-codes.

Some ideas in computing are just hard to put into real world analogies. This is one of those ideas. Trying to apply some sort of IP law to it is just a plain bad idea. People who code know this instinctively, people who don't are just going to have to trust the people who do, or switch the damn things off altogether.