KSL [was Re: Introduction and RFC]

Tom Ball Tom.Ball at Sun.COM
Fri Oct 26 08:15:28 PDT 2007


Stephen Colebourne wrote:
> In addition, KSL has a high barrier of entry (in terms of fulfilling all 
> the many things that compiler writers and language designers should do, 
> including documentation, spec writing and tests).

That's why I like it -- most software developers think they are 
world-class language designers, but the truth is only a very small 
minority are (and I'm not in that group).  Features can be easily added 
to an untyped or loosely-typed language, but most additions to the Java 
language require a rigorous examination of their impact on issues like 
type-safety and security.  Few people take these requirements very 
seriously when throwing out suggestions, but for those who do want the 
level of rigor that working with javac provides, KSL is the place to be. 
  There are other aliases for people who just want to intellectually 
joust; personally, I think those sorts of discussions belong instead in 
a pub with beer for everyone.

> Of course, none of this is wrong, it just may signal that 'we the 
> community' need to create another project separate from KSL where there 
> is a much lower barrier of entry, but as a result a higher risk that the 
> compiler/resulting code will be broken. ie. a place where ideas can 
> actually be investigated.

My guess is that it would be easier to hack BeanShell or Groovy, as 
their type-safety requirements appear to be lower.  I certainly could be 
wrong about that, of course.

Tom

> Stephen
> PS. This discussion about KSL probably should be on the KSL list, but 
> thread-wise it makes sense here for now.
> 
> PPS.I should also note that I originally signed up here as the KSL 
> project webpages told me to do so, so the fact that there is now a KSL 
> mailing list is rather a surprise.
> 
> 
> Frederic Simon wrote:
>> I really thought that the KSL was created exactly for that: Filtering 
>> language proposals. But there are no mailing lists for KSL pure, and 
>> the Bug database should not be populated with KSL noise.
>> About KSL, in my experience adding some code to javac (pure 
>> implementation) to support small language proposal, is a lot faster 
>> and cheaper than:
>> - Evaluate the coherence/readability
>> - Evaluate its usefulness (in writing code and API)
>> - Evaluate its impact on current API
>> - Evaluate the risk impact on the javac and JVM
>>
>> So, the KSL is great. Have it, play with it, throw it away (and "may 
>> be", "sometimes", "rarely", "occasionally": keep it). And it does not 
>> have to be Sun employees doing the steps. For me, once a language 
>> proposal and RFE entry starts to get momentum (votes and so on), so 
>> the team leaders as decided in the GB (Sun for the moment) can get 
>> involved and evaluate the next steps.
>>
>> The KSL for me is  "Extreme Agility", and luxury of having the 
>> implementation before deciding if you need it or not.
>>
>> So, please keep the spirit of it, it's good for everyone.
>>
>> On 10/25/07, *Dalibor Topic* <robilad at kaffe.org 
>> <mailto:robilad at kaffe.org>> wrote:
>>
>>     Ted Neward wrote:
>>      > Interesting--which list *would* be the proper place to discuss
>>     language
>>      > proposals? Is there one? (Personally, I thought it was this one.)
>>      >
>>     I think the kitchen sink language project is the venue you are 
>> looking
>>     for: https://ksl.dev.java.net/
>>
>>     cheers,
>>     dalibor topic
>>
>>
>>
>>
>> -- 
>> http://freddy33.bglogspot.com/ <http://freddy33.bglogspot.com/>
>> http://www.jfrog.org/




More information about the compiler-dev mailing list