[nycbug-talk] eurobsdcon

Bob Ippolito bob
Mon May 24 09:27:35 EDT 2004


On May 24, 2004, at 9:17 AM, Pete Wright wrote:

> Brown, James (Jim) wrote:
>
>>> -----Original Message-----
>>> From: G.Rosamond [mailto:george at sddi.net]
>>> Sent: Sunday, May 23, 2004 10:31 PM
>>> To: Jan Schaumann
>>> Cc: NYC Bug List
>>> Subject: Re: [nycbug-talk] eurobsdcon
>>>
>>> On May 23, 2004, at 10:25 PM, Jan Schaumann wrote:
>>>
>>>
>>>> Dru <dlavigne6 at sympatico.ca> wrote:
>>>>
>>>>> -contributing documentation, tutorials, reviews
>>>>>
>>>> This is one point that can't be stressed enough.  Documentation is
>>>> something that anybody can contribute to - even though
>>> getting really
>>>
>>>> <snip>
>>>>
>>
>> Documentation is important, but there is a more critical need, IMHO.
>>
>> Open Source software needs reliable testing frameworks.  Most Open
>> Source projects of substance are large (> 50K lines of code) and
>> most do not have a reliable, rigorous testing framework. My 
>> perception is that Open Source testing consists mainly of just
>> getting as many people as possible to run the project code
>> and submit bug reports.  While helpful, this is spotty testing.
>>
>>
>> The Perl Test Harness is a step in the right direction.  Test 
>> programs are written for specific features of a module
>> and compared against expected output.  These can be run by
>> anyone, on any hw/sw platform that runs Perl.
>>
>> The BIND 9 test suite is another good example, although in this
>> case, it's not easy to set up and run.
>>
>>
>> Consider challenging the Open Source community to develop testing
>> frameworks for every project.  'make test' should perform a complete 
>> suite of tests for the project that can be reused
>> for every release.
>>
> yea i totally agree with you there.  and hopefully OSDL has started to 
> take on this role, and maybe other organizations will follow.  i do 
> not even know of any opensource software testing out there, is there 
> any?

It's almost exclusively done on a project by project basis.  Testing is 
a lot harder than it sounds.  You more or less have to design your 
software to be easily testable, or you're pretty screwed.  Developing a 
test suite is often just as hard, sometimes harder, than the 
application or library itself.  It takes a much higher level 
understanding of the problem at hand and the implementation of the 
software in order to write effective tests.

-bob





More information about the talk mailing list