Re: How much requirement documentation up front?
From: Phlip (phlip_cpp_at_yahoo.com)
Date: 10/22/04
- Next message: Kevin Cline: "Re: XP Requirement Analysis?"
- Previous message: Sylvain: "How much requirement documentation up front?"
- In reply to: Sylvain: "How much requirement documentation up front?"
- Next in thread: Robert C. Martin: "Re: How much requirement documentation up front?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Fri, 22 Oct 2004 14:30:08 GMT
Sylvain wrote:
> Hi, earlier this week I posted a question regarding RUP and how much spec
> must be done at the end of the Elaboration phase (see below for original
> post). I was quite surprised to see so few replies. I will rephrase my
> question to make it broader and nonn-RUP specific:
> Question: In an iterative development process where formal requirements
are
> written, as opposed to features and user stories, what percentage of the
> requirements need to be formally documented before staffing for the
> construction phase?
The issue is not "in iterative development process". An overwhelming amount
of research of _all_ software development has shown that a large formal
requirements gathering effort correlates with a very high failure rate.
The highly iterative development processes require a simple solution:
Minimize the time between making a decision about specifications, and
collecting real feedback on the results of those decisions.
> Does 80% sound too much for you?
Yes. Do a "spike solution", which is a minimal implementation of the most
important features. Review the result for these criteria:
1) The vision and highest-priority 5% of requirements are stable
2) The architecture, in code, is stable
3) The approach to be used during testing is proven
4) Demonstrate that the technical risks have been deferred or handled
After collecting these data on project topology, add staff, and start adding
more features. The pipeline between making decisions about features and
reviewing results should be one week. That means every possible way to
automate the review should be investigated, including literate acceptance
test frameworks, and automated deployment tools.
> I know there is no magic answer, I just
> want to hear your experience is.
My experience using the miraculous process known as Code-and-Fix, for small
projects, consisted of writing a GUI, getting it reviewed, pushing back into
the features, getting them review, and making sure I never let the "easy
bug" count get too high. So, in retrospect, I was doing things half-right!
> BTW: I also posted my original question to the RUP forum and I got a few
> answers. Follow this link, if your are interested:
>
http://www-128.ibm.com/developerworks/forums/dw_expandTree.jsp?thread=64051&forum=335&cat=24&message=4730759#4730759
They still allow a longer pipeline between decision making and feedback.
(BTW this is a question for news:comp.software-eng ; it's only indirectly
about OO.)
-- Phlip http://industrialxp.org/community/bin/view/Main/TestFirstUserInterfaces
- Next message: Kevin Cline: "Re: XP Requirement Analysis?"
- Previous message: Sylvain: "How much requirement documentation up front?"
- In reply to: Sylvain: "How much requirement documentation up front?"
- Next in thread: Robert C. Martin: "Re: How much requirement documentation up front?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
