Loading ...
Sorry, an error occurred while loading the content.

Re: design practices

Expand Messages
  • woynam
    The key sentence is that all the functionality *may* be needed. I ve seen plenty of features that were never used, but they simply *had* to be there. Often,
    Message 1 of 25 , Jan 5, 2005
    • 0 Attachment
      The key sentence is that all the functionality *may* be needed. I've
      seen plenty of features that were never used, but they simply *had* to
      be there. Often, there's no one who can tell you why a feature is
      needed, or why it was put there in the first place.

      I would start with the highest priority features first. At some point,
      the customer may decide that there's enough functionality to deploy
      into production. Worst case scenario, you rebuild all the features,
      which is basically were you might end up if you used a waterfall
      approach.

      Good case scenario, your customer decides that they can live with a
      subset of the features, and the system gets deployed into production
      eariler. You gain important feedback as you continue developing the
      remaining requested features.

      Best case scenario, the customer decides they *don't* need the
      remaining feqatures, but would rather you work on some *new* feature
      that will benefit the business. If you had taken the waterfall
      approach, your answer may be, "Sorry, we can't do that. We're 12
      months into a 18 month rewrite, and all we have is a bunch of design
      documents. We have to code and test the system first before we can
      begin to worry about changes." :-)

      Mark

      --- In scrumdevelopment@yahoogroups.com, "Joel" <jadams@d...> wrote:
      >
      > I have never managed a re-write project before, so perhaps my
      answer
      > is naive, but here goes anyway.
      > I disagree with some other respondents. I think that you need to
      > modify some Scrum/Agile practices because some assumptions are not
      > valid in this type of project. For example, one assumption of
      Agile
      > is that you should prioritize the functionality because lower
      > priority functionality 1)might never be needed or 2)might not make
      > economic sense by the time we get to it or 3) the user might just
      > change their mind, (or other variations of this theme). One
      drawback
      > of waterfall is that it treats all functionality with equal
      > priority. But, in this case, all functionality may be required.
      The
      > user may not want a replacement system unless it has ALL previous
      > functionality, hence the need for prioritization assumption is no
      > longer valid.
      > Another assumption we frequently make is that design should be
      > emergent. Writing code only for the current function is a good
      thing
      > because re-writing isn't expensive compared to adding complexity
      and
      > functionality that is never needed. But in the case where all
      > functionality has equal priority, and the system cannot be released
      > until all of the functionality is included, isn't this an
      inefficient
      > approach?
      > My guess is that this project would benefit from more up-front
      design
      > than a new development project. And I think that I would be asking
      > the team where else they can find efficiencies given the somewhat
      > unique situation.
      >
      > --- In scrumdevelopment@yahoogroups.com, "Steven Mak" <tcmak@y...>
      > wrote:
      > >
      > > Hi all,
      > >
      > > I am going to have a new project soon and I am the project
      manager
      > of
      > > the project. I want to apply scrum practices with this project. My
      > > knowledge in scrum is limited to this group, some online
      materials,
      > > and a book. Thus, hope this gives me practical experiences in how
      my
      > > development team more 'agile'.
      > >
      > > The project is about re-writing a system that has been using for
      > > serveral years. It's so badly written that it has to be
      re-written
      > so
      > > as to have a easier life to maintain it.
      > >
      > > The first query is not about 'design'. Is traditional software
      > > development works better when we have deep understanding about the
      > > system? (eg. re-writing a system that we have been using for
      > serveral
      > > years) or I would like to know how agile methologies work better
      in
      > a
      > > not-so-emergent system development.
      > >
      > > The scope of the system is rather board, thus I believe it would
      > take
      > > quite a number of iterations before its functionalities can
      compare
      > > with the existing system.
      > >
      > > I am wondering, with such a large system. How infrastruture and
      > > software design is done in Scrum, while not falling into the BDUF
      > > problem? Suppose it's a 4-week iteration. Would 'design' and
      > > 'architecutral' work takes up more time in the first few
      iterations
      > > than the later iterations, as it gets sufficiently-built for most
      > > services on top of it?
      > >
      > > Is there any references on recommended design practices in Scrum?
      > > Thank you.
      > >
      > > Happy Holiday
      > > Steven
    • woynam
      Just a little background on my experience with rewrites. I ve been at my current position for 7+ years, and joined the company to work on a rewrite of our
      Message 2 of 25 , Jan 5, 2005
      • 0 Attachment
        'Just a little background on my experience with rewrites.

        I've been at my current position for 7+ years, and joined the company
        to work on a rewrite of our mainframe-based trading system that
        supported our open-outcry trading model.

        Less than a year into the project, the threat of fully electronic,
        screen-based trading became a reality with the announcement of the
        first new options exchange in 25 years, the ISE. We were soon directed
        to change gears and build a screen-based system.

        Shortly after completing and deploying version 1.0 of the system, the
        trading of single stock futures was authorized, and the fear was that
        a futures exchange could steal a lot of our business. Thus, we began
        developing a screen-based futures trading engine as part of a joint
        venture with 2 other exchanges. This new system was built using 90% of
        the major components of the new system.

        Shortly after completing and deploying the 2nd version of the system,
        the business decided that a hybrid trading model, combining the best
        of open-outcry and screen-based trading, would allow us to better
        compete with the fully electronic markets. Thus, we began working on a
        hybrid trading system that utilizes elements of the mainframe system,
        and well as our new trading platform. We deployed this version over a
        year ago.

        In a nutshell, the mainframe system is still there, and is used for a
        number of products that continue to trade open-outcry, as well as
        being part of the hybrid trading system. Had we focused exclusively on
        replicating the functionality of the original system, we would have
        built the wrong system, as the business did not stand still in the
        face of tough competition. Not only we some features not needed, they
        were completely eliminated with the new business model.

        By iteratively developing and deploying the system, we were able to
        deal with the unexpected course corrections that came our way.

        This is the ultimate value of agility.

        Mark


        --- In scrumdevelopment@yahoogroups.com, "woynam" <woyna@c...> wrote:
        >
        > The key sentence is that all the functionality *may* be needed. I've
        > seen plenty of features that were never used, but they simply *had* to
        > be there. Often, there's no one who can tell you why a feature is
        > needed, or why it was put there in the first place.
        >
        > I would start with the highest priority features first. At some point,
        > the customer may decide that there's enough functionality to deploy
        > into production. Worst case scenario, you rebuild all the features,
        > which is basically were you might end up if you used a waterfall
        > approach.
        >
        > Good case scenario, your customer decides that they can live with a
        > subset of the features, and the system gets deployed into production
        > eariler. You gain important feedback as you continue developing the
        > remaining requested features.
        >
        > Best case scenario, the customer decides they *don't* need the
        > remaining feqatures, but would rather you work on some *new* feature
        > that will benefit the business. If you had taken the waterfall
        > approach, your answer may be, "Sorry, we can't do that. We're 12
        > months into a 18 month rewrite, and all we have is a bunch of design
        > documents. We have to code and test the system first before we can
        > begin to worry about changes." :-)
        >
        > Mark
        >
        > --- In scrumdevelopment@yahoogroups.com, "Joel" <jadams@d...> wrote:
        > >
        > > I have never managed a re-write project before, so perhaps my
        > answer
        > > is naive, but here goes anyway.
        > > I disagree with some other respondents. I think that you need to
        > > modify some Scrum/Agile practices because some assumptions are not
        > > valid in this type of project. For example, one assumption of
        > Agile
        > > is that you should prioritize the functionality because lower
        > > priority functionality 1)might never be needed or 2)might not make
        > > economic sense by the time we get to it or 3) the user might just
        > > change their mind, (or other variations of this theme). One
        > drawback
        > > of waterfall is that it treats all functionality with equal
        > > priority. But, in this case, all functionality may be required.
        > The
        > > user may not want a replacement system unless it has ALL previous
        > > functionality, hence the need for prioritization assumption is no
        > > longer valid.
        > > Another assumption we frequently make is that design should be
        > > emergent. Writing code only for the current function is a good
        > thing
        > > because re-writing isn't expensive compared to adding complexity
        > and
        > > functionality that is never needed. But in the case where all
        > > functionality has equal priority, and the system cannot be released
        > > until all of the functionality is included, isn't this an
        > inefficient
        > > approach?
        > > My guess is that this project would benefit from more up-front
        > design
        > > than a new development project. And I think that I would be asking
        > > the team where else they can find efficiencies given the somewhat
        > > unique situation.
        > >
        > > --- In scrumdevelopment@yahoogroups.com, "Steven Mak" <tcmak@y...>
        > > wrote:
        > > >
        > > > Hi all,
        > > >
        > > > I am going to have a new project soon and I am the project
        > manager
        > > of
        > > > the project. I want to apply scrum practices with this project. My
        > > > knowledge in scrum is limited to this group, some online
        > materials,
        > > > and a book. Thus, hope this gives me practical experiences in how
        > my
        > > > development team more 'agile'.
        > > >
        > > > The project is about re-writing a system that has been using for
        > > > serveral years. It's so badly written that it has to be
        > re-written
        > > so
        > > > as to have a easier life to maintain it.
        > > >
        > > > The first query is not about 'design'. Is traditional software
        > > > development works better when we have deep understanding about the
        > > > system? (eg. re-writing a system that we have been using for
        > > serveral
        > > > years) or I would like to know how agile methologies work better
        > in
        > > a
        > > > not-so-emergent system development.
        > > >
        > > > The scope of the system is rather board, thus I believe it would
        > > take
        > > > quite a number of iterations before its functionalities can
        > compare
        > > > with the existing system.
        > > >
        > > > I am wondering, with such a large system. How infrastruture and
        > > > software design is done in Scrum, while not falling into the BDUF
        > > > problem? Suppose it's a 4-week iteration. Would 'design' and
        > > > 'architecutral' work takes up more time in the first few
        > iterations
        > > > than the later iterations, as it gets sufficiently-built for most
        > > > services on top of it?
        > > >
        > > > Is there any references on recommended design practices in Scrum?
        > > > Thank you.
        > > >
        > > > Happy Holiday
        > > > Steven
      • Steven Gordon
        Emergent design has other benefits over up-front design besides being able to handle changing/uncertain/emergent requirements: 1. Emergent designs force the
        Message 3 of 25 , Jan 5, 2005
        • 0 Attachment
          Emergent design has other benefits over up-front design besides being able to handle changing/uncertain/emergent requirements:

          1. Emergent designs force the code to be easily changable, whereas top-down holistic design can easily create code that is monolithically dependent on the current design. Even if requirements never changed, implementing the system a few requirements at a time results in a system that will be battle tested at being able to accommodate additional requirements after deployment.

          2. Developers can learn something from the design and implementation of a few requirements that will improve the design and implementation of the remaining requirements, whereas if we design the whole thing before implementing any of it, we get no feedback on whether the overall design strategy is a good one.

          3. We can better estimate how long it will take to complete the project from the first month of design and implementation of a few requirements from beginning to end than from just a month of pure design.

          Steven Gordon

          http://sf.asu.edu/




          --- In scrumdevelopment@yahoogroups.com, "Joel" <jadams@d...> wrote:
          > Another assumption we frequently make is that design should be
          > emergent. Writing code only for the current function is a good
          thing
          > because re-writing isn't expensive compared to adding complexity
          and
          > functionality that is never needed. But in the case where all
          > functionality has equal priority, and the system cannot be released
          > until all of the functionality is included, isn't this an
          inefficient
          > approach?
          > My guess is that this project would benefit from more up-front
          design
          > than a new development project. And I think that I would be asking
          > the team where else they can find efficiencies given the somewhat
          > unique situation.
        • todd
          ... This is a bit of FUD. There s nothing in top down design that makes it any more monolithic or dependent or less changeable than any other design approach.
          Message 4 of 25 , Jan 5, 2005
          • 0 Attachment
            Steven Gordon wrote:

            > Emergent design has other benefits over up-front design besides being
            > able to handle changing/uncertain/emergent requirements:
            >
            > 1. Emergent designs force the code to be easily changable, whereas
            > top-down holistic design can easily create code that is monolithically
            > dependent on the current design. Even if requirements never changed,
            > implementing the system a few requirements at a time results in a
            > system that will be battle tested at being able to accommodate
            > additional requirements after deployment.


            This is a bit of FUD. There's nothing in top down design that makes it
            any more monolithic or dependent or less changeable than any other
            design approach. If you know how to design your code to be testable and
            changeable then it will be. And it can be far easier to add in new
            requirements, depending on the nature of the new requirements. If people
            add useless over complicated code that doesn't work then that's there
            own issue, it's not required by any design methodology that i know of.
            Nor will any design methodology prevent someone from doing stupid stuff.
          • Steven Gordon
            The argument that if you know what you are doing, you can create a perfectly good software using any methodology (or even no methodology) defeats anything we
            Message 5 of 25 , Jan 5, 2005
            • 0 Attachment
              The argument that if you know what you are doing, you can create a perfectly good software using any methodology (or even no methodology) defeats anything we say here. Given the current state of software today, this is not a very realistic or productive argument.

              It is common sense that a design that has had new requirements added to it dozens of times in a principled way can be more trusted to be able to accommodate new requirements of similar kinds than a design that has never had any new requirements added to it.

              -----Original Message-----
              From: todd [mailto:todd@...]
              Sent: Wed 1/5/2005 10:53 AM
              To: scrumdevelopment@yahoogroups.com
              Cc:
              Subject: Re: [scrumdevelopment] Re: design practices




              Steven Gordon wrote:

              > Emergent design has other benefits over up-front design besides being
              > able to handle changing/uncertain/emergent requirements:
              >
              > 1. Emergent designs force the code to be easily changable, whereas
              > top-down holistic design can easily create code that is monolithically
              > dependent on the current design. Even if requirements never changed,
              > implementing the system a few requirements at a time results in a
              > system that will be battle tested at being able to accommodate
              > additional requirements after deployment.


              This is a bit of FUD. There's nothing in top down design that makes it
              any more monolithic or dependent or less changeable than any other
              design approach. If you know how to design your code to be testable and
              changeable then it will be. And it can be far easier to add in new
              requirements, depending on the nature of the new requirements. If people
              add useless over complicated code that doesn't work then that's there
              own issue, it's not required by any design methodology that i know of.
              Nor will any design methodology prevent someone from doing stupid stuff.
            • todd
              ... It s productive and realistic because you can t overcome this lack by adding a methodology. So dinging one methodology in favor another isn t fair, imho.
              Message 6 of 25 , Jan 5, 2005
              • 0 Attachment
                Steven Gordon wrote:

                > The argument that if you know what you are doing, you can create a
                > perfectly good software using any methodology (or even no methodology)
                > defeats anything we say here.

                It's productive and realistic because you can't overcome this lack by
                adding a methodology. So dinging one methodology in favor another isn't
                fair, imho.

                >
                > It is common sense that a design that has had new requirements added
                > to it dozens of times in a principled way can be more trusted to be
                > able to accommodate new requirements of similar kinds than a design
                > that has never had any new requirements added to it.

                The use of the word "principled" defeats anything we say here. Given
                the current state of software today, this is not a very realistic or
                productive argument. :-)
              Your message has been successfully submitted and would be delivered to recipients shortly.