test-driven standards

[Steve Loughran <st...@apache.org>]

Dan Diephouse wrote:
> Steve Loughran wrote:
> 
>> On the subject of standards, its my opinion that standards groups need 
>> to adopt a lot more OSS thinking. That means
>>
>> -specs come with tests. Or the tests are the spec, to put it differently
>> -team SCM repos of test+source are public
>> -gump does a  nightly build of everything
>>
>> see: see: 
>> http://cvs.sourceforge.net/viewcvs.py/*checkout*/deployment/deployment/doc/steve/test_driven_standards.doc 
>>
>>
>>
>> I'm partway there with the spec I work on, but there is so much 
>> ignorance (we dont need tests) and unwillingness to put stuff into a 
>> shared SCM that we dont get the full benefits
>>
 > I can't +1 this enough. Is there any kind of legal reason why current
 > JSRs are so closed about their TCKs and sources?
 >


Dont know about legality, but I I kind of hint at other causes in this 
presentation on testing, where I worry about adoption being pushed back 
by management

http://people.apache.org/~stevel/slides/testing.pdf

Interestingly, junit testing is far more prevalent in java OSS than 
'enterprise grade' apps that I often encounter. I havent yet come up 
with a good reason for this, but some hypotheses:

  -pressure from above to deliver code, no management pressure for tests.
  -more innovation in process is happening in OSS
  -in OSS-land, everybody knows not to trust an app without junit tests 
(or regular gump outages)

I have a dream, and the dream is: every W3C standard, every JCP program 
has an SVN repo with a test suite, and gump builds every one, every 
night. Help me realise this dream... (*)


-Steve


(*) I also have a dream that at the click of a button we can refile 
bugreps of the type "Ant doesnt handle '--' in comments" as bugs with 
the XML working group though some distributed bug database, but that one 
is harder,

Re: test-driven standards

[Steve Loughran <st...@apache.org>]

Dan Diephouse wrote:

> 
> I will to try to help realize the dream because it is a dream of mine as 
> well. There is no reason not to make the JSR source and tests available 
> under the CDDL or some such license. I'm not sure how effective I will 
> be through the ASF, but we'll see :-)
> 
> - Dan
> 

Well, I'm trying to overthrow the W3C's process, one TAG member at a 
time. if the real standards bodies can set examples, there is no reason 
why the rest can't.

I guess the problem w/ the JSR groups is that they often view tests as 
strategic things to only be released to people who sign the NDAs, like 
you have to do for TCKs. But in reality they are nothing but a means of 
testing that an implementation meets the design, which is something 
every end user should have the right to do.

-Steve

Re: test-driven standards

[Dan Diephouse <da...@envoisolutions.com>]

Steve Loughran wrote:

> Dan Diephouse wrote:
>
>> Steve Loughran wrote:
>>
>>> On the subject of standards, its my opinion that standards groups 
>>> need to adopt a lot more OSS thinking. That means
>>>
>>> -specs come with tests. Or the tests are the spec, to put it 
>>> differently
>>> -team SCM repos of test+source are public
>>> -gump does a  nightly build of everything
>>>
>>> see: see: 
>>> http://cvs.sourceforge.net/viewcvs.py/*checkout*/deployment/deployment/doc/steve/test_driven_standards.doc 
>>>
>>>
>>>
>>> I'm partway there with the spec I work on, but there is so much 
>>> ignorance (we dont need tests) and unwillingness to put stuff into a 
>>> shared SCM that we dont get the full benefits
>>>
> > I can't +1 this enough. Is there any kind of legal reason why current
> > JSRs are so closed about their TCKs and sources?
> >
>
>
> Dont know about legality, but I I kind of hint at other causes in this 
> presentation on testing, where I worry about adoption being pushed 
> back by management
>
> http://people.apache.org/~stevel/slides/testing.pdf
>
> Interestingly, junit testing is far more prevalent in java OSS than 
> 'enterprise grade' apps that I often encounter. I havent yet come up 
> with a good reason for this, but some hypotheses:
>
>  -pressure from above to deliver code, no management pressure for tests.
>  -more innovation in process is happening in OSS
>  -in OSS-land, everybody knows not to trust an app without junit tests 
> (or regular gump outages)
>
> I have a dream, and the dream is: every W3C standard, every JCP 
> program has an SVN repo with a test suite, and gump builds every one, 
> every night. Help me realise this dream... (*)
>
Already read your presentation from your blog, but thanks. I totally 
agree with you. And FWIW your experience with unit testing in the 
enterprise lines up completely with mine. I for the most part feel its 
ingorance. A lot of organizations are isolated and live in their own 
little world.

I will to try to help realize the dream because it is a dream of mine as 
well. There is no reason not to make the JSR source and tests available 
under the CDDL or some such license. I'm not sure how effective I will 
be through the ASF, but we'll see :-)

- Dan

-- 
Dan Diephouse
Envoi Solutions LLC
http://netzooid.com