RE: [XP] Re: [scrumdevelopment] Re: [APM] Re: Refactoring Requirements over Tracing them (was RE: Business Traceability)
- On Wed, Mar 17, 2004 at 07:47:09PM -0600, Gary Brown wrote:
> This is just crap. I have been watching this thread closely, what IBrad Appleton wrote:
> see are a lot of folks who want to add a bunch of useless tasks to
> Agile/XP projects.
> Whoa! Who are you claiming has said they WANT to add this stuff? IUnfortunately, this is true. As the government mandates how
> certainly haven't. I started from the very beginning saying I think
> traceability is predictive rather than adaptive. Then I (and even
> Mike) also acknowledged that bunches of folks are in the boat of
> having to meet a traceability mandate for their projects (not
> something they WANT, something that is being mandated of them) and
> that, for better or worse, SOX is likely to increase this trend (some
> are saying it already has), again NOT because it is legally required
> by SOX but because of fear
to undergo business activities, traceability will come down
to systems through a variety of regulations ranging from
"financial controls" (SOX), to "necessity" and/or 'procedural"
Gary Brown wrote:
> All of this can be captured at the story level by the appropriate SCMBrad Appleton wrote:
> I think that is EXACTLY what I have been arguing (thank you forYes, I do say this, but I think the system you propose
> reiterating that). And I think Mike tried to argue that would be
> insufficient cuz it would miss stuff, and wouldn't catch everything.
> And I argued I don't need perfection/everything, just "good enough",
> and I think the above does that.
"has value" -- like I said, I would love to download
it and use it.
Gary Brown wrote:
> Executable tests are better than written documentation. TheBrad Appleton wrote:
> test are always up to date. The documentation usually
> is not, in my experience.
> No one here that I know has been arguing against thatWell, I thought you did ;-), but I am glad to hear you don't,
> either. No one here has argued for docs instead of tests,
> nor for traceability instead of tests.
> Nonetheless, there are going to be those which will have aBrad, you probably missed a section of one of my posts.
> traceability mandate to meet and will want to be as agile as they can
> be while doing the simplest thing possible to meet that mandate. And
> I've claimed if you DO have such a mandate, executable tests alone
> won't meet it (if it did, then you really don't have such a mandate).
But in the unfortunate event that you are required or
forced to do some traceability (yuck!), I think *you can*
generate lots of traceability data through the
This is a proposal similar but different than you
"SCM-based automated traceability".
Here is how it would work:
1) Write every acceptance test for a story
in a way that the logger dumps a "dependent trace",
spitting out classes as they are used in the
2) Parse the outcome for each acceptance test
to understand the *dependency list* of every
3) Integrate (combine the dependency lists)
across all acceptance tests for all stories.
You can then turn on or off this tracing mechanism in
production, but "the system" can generate its own
"traceability matrix", I think,
"A man who has the knowledge but lacks the power clearly to
express it is no better off than if he never had any ideas at all."