Hacker News Reader

Comment by Aqueous | Hacker News Reader

I actually think "Typesafe" is an excellent name for a company - whether or not it has grown to be more than just Scala. It says to me: If you build on top of our software, you will be safe and protected due to our built-in guarantees. That is a good foundational brand identity for any company. I don't think you should change it.

View on HN

My first thought. I looked at their logo and then was expecting something in the lines of gettypesafe.com, but was pleasantly surprised that they own that domain.

Typesafe is an excellent name, and it sounds very professional for they space they're in.

I only skimmed the article but it seemed they were starting to deal more in softwares that are not typesafe, therefore the name is misleading. It's a great name for a Haskell shop!

What they say is that a lot of their clients use Java (since most of the products have a dual Java/Scala API) and Java is not the most typesafe language there is.

Also Akka is not very typesafe, with the actors message being of the "Any" type!

That's misleading. Yes, you can pass anything as a message, but you pattern match the message to find the type. I suppose you could typecast the message, but, uh... why would you ever do that?

How do you get compile-time errors, which are the point of statically typed languages?

They're currently working on "typed actors" which will hopfully absolve the Scala world of these awful TypeDangerous actors, thus redeeming the name "TypeSafe"

If you cast, you don't. If you pattern match, you do. Pattern matching allows you to determine the type of an object and extract its contents, by "testing" which type it is. Everything is typechecked.

The compiler prevents you from doing

  case x: Warble => modifyTheWargle(x)

but it doesn't stop you from sending out a Wargle to the actor that only expects Warbles.

I know what pattern matching is; I'm not unfamiliar with Scala or other statically-typed languages with pattern matching. I was referring to what user morsch is saying in his/her comment.

The problem I see is in the sender, who doesn't know if the message makes sense for the receiver. Isn't this like programming in Smalltalk or Ruby, with their "doesNotUnderstand" or "method_missing"? How does this mesh with Scala's static typing, which aims to give you compile-time errors when you screw up?

Interesting, can you name a distributed system that preserves types across machines?

Other than the experimental typed actors library that Akka is working on, and Cloud Haskell, I'm not aware of any type preserving distributed system, certainly nothing production ready/widely used.

So, the critique, "oh no, Any, how horrible!" seems absurd given the current state-of-the-art ;-)

The word 'Jini' is bubbling up from the depths of my memory... Java-based distributed architecture which relied on the 'write once run anywhere' model to distribute proxies for remote objects...

I like the name as well. Apart from the meaning of the name to engineers it's easy to remember because its constituent parts are very common English words.

I can imagine though that if you want to offer reactive systems with coast-to-coast backpressure you perhaps want to support parts of the system being in, say, Javascript as well.

I'd seriously consider keeping the name and just treating it no longer as a word but purely as a name. I.e. something which dictionary definition is not constraining to the company mission.

Agreed.

I worry that the name change will reflect a change in focus away from type-safety. Which, like someone else commented, is already the case with (some aspects of) Akka and UntypedActors.

Reply on news.ycombinator.com