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

Re: User-stories for components

Expand Messages
  • Rhydian Thomas
    Hi Steve, In a nutshell ... the product was a framework which the customer team would take and build a UI on top of and implement various integration points
    Message 1 of 27 , Mar 4, 2009
    • 0 Attachment
      Hi Steve,

      In a nutshell ... the 'product' was a framework which the customer team would take and build a UI on top of and implement various integration points (e.g. with billing servers, content management systems, etc) ... So our aim was to develop and maintain the various components of this framework such that it could accomodate the requirements of the customer team.

      We used Scrum, development was incremental and iterative, in each iteration we would tackle bugs, feature enhancements, etc. which centred around different component parts.

      In that context, I wonder whether user-stories could have been used as a requirements tool. Can you use the same concepts of a short narrative to capture the feature, discussion with business owners to ellaborate on detail, acceptance criteria to know when we're done and automate them ... when there is no UI, no real customer.



      --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@...> wrote:
      >
      > On Wed, Mar 4, 2009 at 2:54 PM, Rhydian Thomas <rhydthom@...> wrote:
      > >
      > > On 05/03/2009, at 8:38 AM, Steven Gordon wrote:
      > >> Many places I have seen component teams, they fear getting behind so
      > >> much that they start including anticipated future functionality in
      > >> everything they do. In the best case, there is just a bunch of YAGNI;
      > >> in the worst case. it turns into churning out widgets and telling the
      > >> application teams that the widgets provide everything they should ever
      > >> need to complete their stories.
      > >>
      > >> A common factor for this worst case is in the context of product
      > >> lines, where the component teams often take on the mission of
      > >> facilitating interoperability and product consistency by mandating
      > >> what widgets the application teams will use and then proceed to build
      > >> them whether or not they support the stories the application teams are
      > >> currently working on.
      > >>
      > >
      > > That sounds like the model being used at the company I worked for when
      > > developing the example component I mentioned in my original post -
      > > product lines, goal to build 'flexible', configurable, reusable
      > > components that could be pieced together and configured when another
      > > customer came along. So perhaps I was struggling to see how user-
      > > stories could be used as I was attempting to use an agile practice in
      > > a non-agile environment? Can agile work in such an environment?
      > >
      >
      > Do you mean user stories in an otherwise non-agile approach? I am not
      > sure if that can accomplish much.
      >
      > Do you mean every team iteratively delivers on a backlog, but the
      > backlog for the architectural teams are the widgets that would seem to
      > be generally useful for the user stories on the application teams'
      > backlogs? It can work, just not as well as if the architectural teams
      > directly collaborate on the application-level user stories. I would
      > generally expect more bloat, more architecture that goes unused, more
      > code complexity, a higher cost of maintenance, a higher cost of change
      > and a slower responsiveness to change than there should have to be.
      >
      > Engineer utilization would likely be higher under this approach. But,
      > is the goal to keep the engineers busy or minimize how long it takes
      > to deliver features? This is a fallacy that project management falls
      > into because a Gant Chart tells you that maximizing utilization will
      > decrease how long it takes to deliver the whole project.
      >
      > Is the goal of a server farm to maximize utilization of the servers or
      > minimize transaction times? Is the goal of building a new freeway to
      > maximize the number of cars that use it or minimize how long it takes
      > people to go from point A to point B? For some reason, people see the
      > fallacy of maximizing utilization much better with inanimate objects
      > than with people.
      >
      > Steve
      >
    • Steven Gordon
      ... The product is a framework and the real customers are development teams in different organizations around the world that would be using the framework to
      Message 2 of 27 , Mar 4, 2009
      • 0 Attachment
        On Wed, Mar 4, 2009 at 6:21 PM, Rhydian Thomas <rhydthom@...> wrote:
        > Hi Steve,
        >
        > In a nutshell ... the 'product' was a framework which the customer team
        > would take and build a UI on top of and implement various integration points
        > (e.g. with billing servers, content management systems, etc) ... So our aim
        > was to develop and maintain the various components of this framework such
        > that it could accomodate the requirements of the customer team.
        >
        > We used Scrum, development was incremental and iterative, in each iteration
        > we would tackle bugs, feature enhancements, etc. which centred around
        > different component parts.
        >
        > In that context, I wonder whether user-stories could have been used as a
        > requirements tool. Can you use the same concepts of a short narrative to
        > capture the feature, discussion with business owners to ellaborate on
        > detail, acceptance criteria to know when we're done and automate them ...
        > when there is no UI, no real customer.
        >

        The product is a framework and the real customers are development
        teams in different organizations around the world that would be using
        the framework to build whatever their customer's need. Right?

        To do agile, you would still need a "customer" (i.e., product owner in
        Scrum) to create user stories that adequately represent what the
        actual customers need and prioritize them. I am sure it is possible
        to write appropriate stories.

        You also need some way for that customer to determine that those
        stories have been implemented acceptably. That would seem to be
        difficult without actually utilizing the functionality in the way you
        would expect a real customer to utilize it.

        I would strongly suggest that there be a team that builds a typical
        customer application using the framework product to not only verify
        that the framework really does what it is supposed to, but also to
        provide customers with the code of a nice working example of how to
        use the framework (like the Pet Store app that came along with early
        releases of J2EE - does it still come with it?). The more coherent,
        automated test code that comes with the framework and the sample app,
        the better for your customers.

        Steve
      • jmilunsky
        It really doesn t matter whether you re delivering UI to the customer or not. User stories are still the way to go to manage the requirements process. You
        Message 3 of 27 , Mar 4, 2009
        • 0 Attachment
          It really doesn't matter whether you're delivering UI to the customer or not. User stories are still the way to go to manage the requirements process. You should check out Mike Cohn's site. He has a lot of information about Epics, themes, and stories and how to break them down. His site is mountaingoatsoftware.com

          Hope this helps
          Jack
          www.agilebuddy.com
          blog.agilebuddy.com

          --- In extremeprogramming@yahoogroups.com, "Rhydian Thomas" <rhydthom@...> wrote:
          >
          > Hi Steve,
          >
          > In a nutshell ... the 'product' was a framework which the customer team would take and build a UI on top of and implement various integration points (e.g. with billing servers, content management systems, etc) ... So our aim was to develop and maintain the various components of this framework such that it could accomodate the requirements of the customer team.
          >
          > We used Scrum, development was incremental and iterative, in each iteration we would tackle bugs, feature enhancements, etc. which centred around different component parts.
          >
          > In that context, I wonder whether user-stories could have been used as a requirements tool. Can you use the same concepts of a short narrative to capture the feature, discussion with business owners to ellaborate on detail, acceptance criteria to know when we're done and automate them ... when there is no UI, no real customer.
          >
          >
          >
          > --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@> wrote:
          > >
          > > On Wed, Mar 4, 2009 at 2:54 PM, Rhydian Thomas <rhydthom@> wrote:
          > > >
          > > > On 05/03/2009, at 8:38 AM, Steven Gordon wrote:
          > > >> Many places I have seen component teams, they fear getting behind so
          > > >> much that they start including anticipated future functionality in
          > > >> everything they do. In the best case, there is just a bunch of YAGNI;
          > > >> in the worst case. it turns into churning out widgets and telling the
          > > >> application teams that the widgets provide everything they should ever
          > > >> need to complete their stories.
          > > >>
          > > >> A common factor for this worst case is in the context of product
          > > >> lines, where the component teams often take on the mission of
          > > >> facilitating interoperability and product consistency by mandating
          > > >> what widgets the application teams will use and then proceed to build
          > > >> them whether or not they support the stories the application teams are
          > > >> currently working on.
          > > >>
          > > >
          > > > That sounds like the model being used at the company I worked for when
          > > > developing the example component I mentioned in my original post -
          > > > product lines, goal to build 'flexible', configurable, reusable
          > > > components that could be pieced together and configured when another
          > > > customer came along. So perhaps I was struggling to see how user-
          > > > stories could be used as I was attempting to use an agile practice in
          > > > a non-agile environment? Can agile work in such an environment?
          > > >
          > >
          > > Do you mean user stories in an otherwise non-agile approach? I am not
          > > sure if that can accomplish much.
          > >
          > > Do you mean every team iteratively delivers on a backlog, but the
          > > backlog for the architectural teams are the widgets that would seem to
          > > be generally useful for the user stories on the application teams'
          > > backlogs? It can work, just not as well as if the architectural teams
          > > directly collaborate on the application-level user stories. I would
          > > generally expect more bloat, more architecture that goes unused, more
          > > code complexity, a higher cost of maintenance, a higher cost of change
          > > and a slower responsiveness to change than there should have to be.
          > >
          > > Engineer utilization would likely be higher under this approach. But,
          > > is the goal to keep the engineers busy or minimize how long it takes
          > > to deliver features? This is a fallacy that project management falls
          > > into because a Gant Chart tells you that maximizing utilization will
          > > decrease how long it takes to deliver the whole project.
          > >
          > > Is the goal of a server farm to maximize utilization of the servers or
          > > minimize transaction times? Is the goal of building a new freeway to
          > > maximize the number of cars that use it or minimize how long it takes
          > > people to go from point A to point B? For some reason, people see the
          > > fallacy of maximizing utilization much better with inanimate objects
          > > than with people.
          > >
          > > Steve
          > >
          >
        • Kim Gräsman
          Torbjörn, On Wed, Mar 4, 2009 at 23:12, Torbjörn Gyllebring ... Right, this is precisely what I meant earlier, thanks for reaffirming :-) ... Good point,
          Message 4 of 27 , Mar 4, 2009
          • 0 Attachment
            Torbjörn,

            On Wed, Mar 4, 2009 at 23:12, Torbjörn Gyllebring
            <torbjorn.gyllebring@...> wrote:
            >
            > I should have been clearer with how I envisioned it based upon Kim's
            > initial post.
            > What I wanted to say was that I see no harm in having a component
            > "team" that is scattered between all clients of that component. Let's
            > call them specialists, guardians or evangelists their title really
            > doesn't matter for this discussion, although depending on your culture
            > you should probably choose carefully. As team members for their
            > vertical teams theese individuals are nothing but members with a
            > stronger inclination for that component, no different from Joe the UI
            > dude or Jane the bit-twiddling-sorcess. So Ivan the Infrastructure
            > guide probably likes that sort of work, will probably sign up for to
            > pair on most of them and as such will in that team be steering the
            > infrastructure work. Ivan also meets on an "as-needed" basis with the
            > other members of the infrastructure group to coordinate bigger changes
            > and get a feel for where the rest of the organization is moving, this
            > is in essence a Scrum of Scrums for a specific product, the shared
            > infrastructure, with some close by and really picky customers, all
            > other developers.

            Right, this is precisely what I meant earlier, thanks for reaffirming :-)

            > Yes if it turns into a dedicated "we do nothing but infrastructure
            > work all the time" clique I see a real danger but in the rosy scenario
            > above things should be fine. If big chunks of functioanlity needs to
            > be developed and can't be partioned, iterated, broken down and built
            > by the vertical teams then I guess some sort of rotation scheme could
            > be established to get create a temporary ad-hoc component-only team
            > for some duration.

            Good point, thanks!

            Oh, and thanks, Steven, for expounding further.

            - Kim
          • tony_t_tubbs
            ... I ve been wondering about these same things myself, and am enjoying this thread very much. My experience is also with large development organizations with
            Message 5 of 27 , Mar 5, 2009
            • 0 Attachment
              --- In extremeprogramming@yahoogroups.com, Steven Gordon
              <sgordonphd@...> wrote:
              >
              > Yes, there may be no perfect solution. So, perhaps a hybrid model
              might work:
              >
              > 1. Have the stories driven vertically by verticals teams.
              > 2. Maintain additional small teams of infrastructural component
              > specialists (guardians).
              > 3. For each task of a vertically driven story that involves extending
              > or modifying a specific infrastructural component, a member of the
              > vertical team pairs with a member of the guardian team for that
              > component.
              >
              > When component team members are idle, they could do one of the
              > following depending on priorities:
              > - refactoring of their component's code,
              > - improvement of the performance of their component (without changing
              > functionality),
              > - help on any vertical stories outside of their component.
              >
              > Steve

              I've been wondering about these same things myself, and am enjoying this
              thread very much. My experience is also with large development
              organizations with many teams, and while reading various agile thoughts
              I often have the same 'in the real world' thoughts that start flames in
              these groups. If I may, I'd like to get down to some basic
              understanding of this whole vertical slice teams idea. I just can't
              seem to see how that could work. My vast ~15 years of experience fits
              this description:

              The organization develops multiple products all related to the same
              industry. For example financial related could include products dealing
              with stocks, bonds, mutual funds, retirement plans, etc or medical
              related could deal with billing, claims processing, insurance
              verification, verification of inseparability, etc. These products
              include not only the 1st order applications, but the printing and
              mailing systems, job queuing systems, etc. Some products produced have
              a narrow audience (HR department) and others are larger (call center
              phone reps). This big picture is then subdivided into teams on other
              continents dealing with international business, and teams in various
              states for US business. Each one of these divisions is responsible for
              multiple products that they develop. The clients use a mix of products,
              and this mix can and often does come from different teams. Though I've
              found it has always been a struggle, the organization as well as the
              clients want these products to easily talk to each other and look and
              feel the same / very similar. This has lead to some specialized teams,
              such as an infrastructure team and a human factors team, teams that are
              trying to ensure consistency across the organization but are not
              typically part of the front-end development teams. The actual
              development teams also have a division of labor between the desktop(PC,
              web, etc) and the SQL/COBOL guys who do their stuff on the mainframe.
              There's also a group everyone works with that controls the database and
              is there to ensure its integrity. It is my expreienc that this team is
              the heaviest bottle neck in that everyone else needs access to the data
              to build their apps. In some environments it is often not possible to
              schedule a meeting with them with less than a months notice. Human
              factors folks often come off to development types as if they are all
              theory, and have no practical experience. I've tended to agree, but am
              not sure that is exactly a fair representation. I am sure much of the
              problem is that their design tools are not the same tools developers
              implement with (Design in Visio and develop in HTML and JavaScript for
              example). The infrastructure team members often get loaned out to help
              at least jump start the front-end development, and we (yes I'm on the
              Infra side) have tried to make the components we produce work like
              internal open source so that we could pair with teams or teams could
              enhance things on there own so that we do not become the bottle-neck,
              but we have not realized this. Most teams still wait on us (they have
              other tasks they could be working on), and, in my probably biased
              opinion, use it as something to point fingers at.

              So, my questions:
              - Does this vertical slice idea mean developers must know both the PC
              languages/tools as well as the mainframe? Have you had success finding
              people that want to do both, or are at least fine with doing so? My
              experience is that there are folks on each side of that that have no
              desire to learn the other.

              - How would you restructure to get the vertical slicing achieved, or
              even just to start it on the path to agile development? It seems
              natural that a separate Infra, DB and Usability teams have emerged out
              of this, I need help seeing how that could become different.

              - If you were to reverse things as it seems to be the implication from
              this tread, and each team were to do slices themselves, including the
              components they use, what would be the technique to ensure the cross
              application communication, similar look and feel, and all the DB science
              (fast indexing, normalization, etc) are achieved? (Basically, the
              desire is that all apps coming from the organization look and feel as if
              they came from the same organization.)










              [Non-text portions of this message have been removed]
            • Steven Gordon
              ... Tony, This is a problem. Consider the following approach where the components have such different technologies: 1. Vertical teams consist of testers and
              Message 6 of 27 , Mar 5, 2009
              • 0 Attachment
                On Thu, Mar 5, 2009 at 8:15 AM, tony_t_tubbs <tony_t_tubbs@...> wrote:
                >
                >
                > --- In extremeprogramming@yahoogroups.com, Steven Gordon
                > <sgordonphd@...> wrote:
                >>
                >> Yes, there may be no perfect solution. So, perhaps a hybrid model
                > might work:
                >>
                >> 1. Have the stories driven vertically by verticals teams.
                >> 2. Maintain additional small teams of infrastructural component
                >> specialists (guardians).
                >> 3. For each task of a vertically driven story that involves extending
                >> or modifying a specific infrastructural component, a member of the
                >> vertical team pairs with a member of the guardian team for that
                >> component.
                >>
                >> When component team members are idle, they could do one of the
                >> following depending on priorities:
                >> - refactoring of their component's code,
                >> - improvement of the performance of their component (without changing
                >> functionality),
                >> - help on any vertical stories outside of their component.
                >>
                >> Steve
                >

                >
                > So, my questions:
                > - Does this vertical slice idea mean developers must know both the PC
                > languages/tools as well as the mainframe? Have you had success finding
                > people that want to do both, or are at least fine with doing so? My
                > experience is that there are folks on each side of that that have no
                > desire to learn the other.

                Tony,

                This is a problem.

                Consider the following approach where the components have such
                different technologies:
                1. Vertical teams consist of testers and the generalists (like myself)
                who do like working in multiple technologies to make sure things
                really work right,
                2. Horizontal (component) teams consist of specialists in each
                specific technology,
                3. Vertical team members implement vertical user stories by
                collaborating on specific tasks with appropriate horizontal team
                members.

                This way the vertical stories are the scheduling and planning
                mechanism to deliver the highest priority features within iteration
                boundaries so as to reduce miscommunication and feedback latency
                times.

                >
                > - How would you restructure to get the vertical slicing achieved, or
                > even just to start it on the path to agile development? It seems
                > natural that a separate Infra, DB and Usability teams have emerged out
                > of this, I need help seeing how that could become different.

                People on these teams would collaborate on vertical stories as
                coordinated by the vertical teams.

                >
                > - If you were to reverse things as it seems to be the implication from
                > this tread, and each team were to do slices themselves, including the
                > components they use, what would be the technique to ensure the cross
                > application communication, similar look and feel, and all the DB science
                > (fast indexing, normalization, etc) are achieved? (Basically, the
                > desire is that all apps coming from the organization look and feel as if
                > they came from the same organization.)
                >

                The component specialists collaborating on the vertical stories should
                be responsible for knowing about these constraints, ensuring they are
                met by any changes or new development they are involved in, and
                perhaps scheduling intense optimization work to be done when no
                vertical team is asking for their time.

                ==============================================================================

                The key point is that we do not just hand off work to the specialists,
                but have generalists who are responsible for delivering working
                vertical slices actively collaborate on each piece with specialists.
                After the stories work acceptably with sufficient automated tests to
                make sure they continue to do so, the specialists can refine the code
                for performance and consistency when that is their highest priority
                tasks.

                This is not to say this approach does not have problems that would
                have to be worked through. I know of no organization that uses such
                an approach.

                Steve
              • tony_t_tubbs
                ... Are you suggesting that the front-end development teams that are currently composed of essentially the UI/View developers (PC,Web folks), and the data
                Message 7 of 27 , Mar 5, 2009
                • 0 Attachment
                  --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@...> wrote:
                  > Consider the following approach where the components have such
                  > different technologies:
                  > 1. Vertical teams consist of testers and the generalists (like myself)
                  > who do like working in multiple technologies to make sure things
                  > really work right,
                  > 2. Horizontal (component) teams consist of specialists in each
                  > specific technology,
                  > 3. Vertical team members implement vertical user stories by
                  > collaborating on specific tasks with appropriate horizontal team
                  > members.
                  >
                  > This way the vertical stories are the scheduling and planning
                  > mechanism to deliver the highest priority features within iteration
                  > boundaries so as to reduce miscommunication and feedback latency
                  > times.

                  Are you suggesting that the front-end development teams that are currently composed of essentially the UI/View developers (PC,Web folks), and the data access folks (COBOL/Mainframe) are restructured so that all members of the team code in all layers? I know only a few people who can and like doing so. It seems the PC guys have a strong distaste for doing mainframe work and vise versa. You say "Horizontal (component) teams" above, but I'm not clear who you are referring to by that statement. This seems to be in contrast to the generalists you mention (which I'm thinking means codes in all layers). So, I am not understanding what happens to the UI and data access folks I just mentioned, as they are a tightly coupled team in that the data provided is intended for that UI being built, but they are not generalist nor component folks. Are they rerouted into other teams, or maybe forced (tactfully) into becoming generalists?. Now the infra teams I mentioned I see the component nature of their work (widgets, file uploaders, system loggers, etc.), but I don't see how that applies to the front end guys.

                  If it is the Human Factors, Infra, and DB teams that are what you called the horizontal teams, then I understand what you've said and find your suggestions do make sense to me. If your comments include the idea of somehow the front end teams have horizontal folks on them, I'm confused.


                  > This is not to say this approach does not have problems that would
                  > have to be worked through. I know of no organization that uses such
                  > an approach.

                  You mean the approach you've suggested, right? Not the one I've described?
                • Steven Gordon
                  I am not sure how to communicate my proposal more clearly. Do you have the assumption that software development work should always be done by a individual
                  Message 8 of 27 , Mar 5, 2009
                  • 0 Attachment
                    I am not sure how to communicate my proposal more clearly.

                    Do you have the assumption that software development work should
                    always be done by a individual person working by themselves in their
                    little cubicle?

                    Is it not possible for two poeple to sit together and complete a
                    software development subtask of a story, where:
                    - the generalist has the understanding of the vertical story and how
                    all the components should work together to accomplish that story, and
                    - the specialist has the deep understanding of the particular
                    component involved with the specific task and any consistency
                    constraints for that component?

                    http://www.google.com/search?q=pair+programming

                    On Thu, Mar 5, 2009 at 10:03 AM, tony_t_tubbs <tony_t_tubbs@...> wrote:
                    > --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@...>
                    > wrote:
                    >> Consider the following approach where the components have such
                    >> different technologies:
                    >> 1. Vertical teams consist of testers and the generalists (like myself)
                    >> who do like working in multiple technologies to make sure things
                    >> really work right,
                    >> 2. Horizontal (component) teams consist of specialists in each
                    >> specific technology,
                    >> 3. Vertical team members implement vertical user stories by
                    >> collaborating on specific tasks with appropriate horizontal team
                    >> members.
                    >>
                    >> This way the vertical stories are the scheduling and planning
                    >> mechanism to deliver the highest priority features within iteration
                    >> boundaries so as to reduce miscommunication and feedback latency
                    >> times.
                    >
                    > Are you suggesting that the front-end development teams that are currently
                    > composed of essentially the UI/View developers (PC,Web folks), and the data
                    > access folks (COBOL/Mainframe) are restructured so that all members of the
                    > team code in all layers? I know only a few people who can and like doing so.
                    > It seems the PC guys have a strong distaste for doing mainframe work and
                    > vise versa. You say "Horizontal (component) teams" above, but I'm not clear
                    > who you are referring to by that statement. This seems to be in contrast to
                    > the generalists you mention (which I'm thinking means codes in all layers).
                    > So, I am not understanding what happens to the UI and data access folks I
                    > just mentioned, as they are a tightly coupled team in that the data provided
                    > is intended for that UI being built, but they are not generalist nor
                    > component folks. Are they rerouted into other teams, or maybe forced
                    > (tactfully) into becoming generalists?. Now the infra teams I mentioned I
                    > see the component nature of their work (widgets, file uploaders, system
                    > loggers, etc.), but I don't see how that applies to the front end guys.
                    >
                    > If it is the Human Factors, Infra, and DB teams that are what you called the
                    > horizontal teams, then I understand what you've said and find your
                    > suggestions do make sense to me. If your comments include the idea of
                    > somehow the front end teams have horizontal folks on them, I'm confused.
                    >
                    >> This is not to say this approach does not have problems that would
                    >> have to be worked through. I know of no organization that uses such
                    >> an approach.
                    >
                    > You mean the approach you've suggested, right? Not the one I've described?
                    >
                  • tony_t_tubbs
                    ... No ... Yes, that is possible, and is being done. This helps clarify for me. The use of specialist in this paragraph implies an OO or data access
                    Message 9 of 27 , Mar 5, 2009
                    • 0 Attachment
                      --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@...> wrote:
                      >
                      > I am not sure how to communicate my proposal more clearly.
                      >
                      > Do you have the assumption that software development work should
                      > always be done by a individual person working by themselves in their
                      > little cubicle?

                      No

                      > Is it not possible for two poeple to sit together and complete a
                      > software development subtask of a story, where:
                      > - the generalist has the understanding of the vertical story and how
                      > all the components should work together to accomplish that story, and
                      > - the specialist has the deep understanding of the particular
                      > component involved with the specific task and any consistency
                      > constraints for that component?

                      Yes, that is possible, and is being done. This helps clarify for me. The use of 'specialist' in this paragraph implies an OO or data access specialist. I took the term in the previous post to refer to be Infrastructure, DBAs and Usability only. With this new understanding, it seems to me there are two different and distinct types of specialists. There are the in-team specialist (OO folks and Data folks, and I guess the generalists we don't really have now) who make up the team proper, are focused on a particular product (or three), and live the life of the project. Then there are the out-of-team specialists that have the focus of corporate consistency stuff, BUT instead of coding and throwing components over the wall, they are brought in as needed and pair with the application team (which implies they will need to pair with multiple teams). We have much room for improvement here, but we do have some level of pairing going on. Typically the data folks and the DBAs do some pairing, and the OO folks and us Infra widget makers do some work together. Granted, most of this is done as a sidebar to the product which results in something like a paired approach to throwing components over the wall, but is a start.

                      Thanks again
                    • Steven Gordon
                      ... Glad you get my meaning now. The whole point is to try to find a way to avoid throwing stuff over the wall, because throwing stuff over the wall makes for:
                      Message 10 of 27 , Mar 5, 2009
                      • 0 Attachment
                        On Thu, Mar 5, 2009 at 11:54 AM, tony_t_tubbs <tony_t_tubbs@...> wrote:
                        > --- In extremeprogramming@yahoogroups.com, Steven Gordon <sgordonphd@...>
                        > wrote:
                        >>
                        >> I am not sure how to communicate my proposal more clearly.
                        >>
                        >> Do you have the assumption that software development work should
                        >> always be done by a individual person working by themselves in their
                        >> little cubicle?
                        >
                        > No
                        >
                        >> Is it not possible for two poeple to sit together and complete a
                        >> software development subtask of a story, where:
                        >> - the generalist has the understanding of the vertical story and how
                        >> all the components should work together to accomplish that story, and
                        >> - the specialist has the deep understanding of the particular
                        >> component involved with the specific task and any consistency
                        >> constraints for that component?
                        >
                        > Yes, that is possible, and is being done. This helps clarify for me. The use
                        > of 'specialist' in this paragraph implies an OO or data access specialist. I
                        > took the term in the previous post to refer to be Infrastructure, DBAs and
                        > Usability only. With this new understanding, it seems to me there are two
                        > different and distinct types of specialists. There are the in-team
                        > specialist (OO folks and Data folks, and I guess the generalists we don't
                        > really have now) who make up the team proper, are focused on a particular
                        > product (or three), and live the life of the project. Then there are the
                        > out-of-team specialists that have the focus of corporate consistency stuff,
                        > BUT instead of coding and throwing components over the wall, they are
                        > brought in as needed and pair with the application team (which implies they
                        > will need to pair with multiple teams). We have much room for improvement
                        > here, but we do have some level of pairing going on. Typically the data
                        > folks and the DBAs do some pairing, and the OO folks and us Infra widget
                        > makers do some work together. Granted, most of this is done as a sidebar to
                        > the product which results in something like a paired approach to throwing
                        > components over the wall, but is a start.
                        >
                        > Thanks again
                        >

                        Glad you get my meaning now.

                        The whole point is to try to find a way to avoid throwing stuff over
                        the wall, because throwing stuff over the wall makes for:
                        - miscommunication that lead to incorrect work and eventual rework.
                        - building more functionality than is actually needed,
                        - long chains of feedback loops instead of time-bounded, direct feedback loops.

                        The organizational resistance is usually based on a specialized
                        division of labor being a more efficient use of resources. If it were
                        not for the 3 problems above, this might well be true.

                        Please, understand that I have never seen an organization implement
                        this approach. I do think it could be made to work well.

                        Steve
                      • MasaKevin Maeda
                        When component team members are idle, they could do one of the following depending on priorities: - refactoring of their component s code, - improvement of
                        Message 11 of 27 , Mar 5, 2009
                        • 0 Attachment
                          "When component team members are idle, they could do one of the

                          following depending on priorities:

                          - refactoring of their component's code,

                          - improvement of the performance of their component (without changing

                          functionality) ,

                          - help on any vertical stories outside of their component."

                          "One solution is to get extremely formal, spec everything out in detail
                          and coordinate a schedule of the work each component team does, but
                          that is clearly not in the spirit of Agile. Short of that, if
                          different teams are doing different parts of the work at different
                          times, there will be different interpretations of what that work was
                          supposed to do."

                          I would add to the two statements above that I would encourage the team to do more testing before helping on vertical stories outside of their component; and if testing is done right then those tests result in more detailed documentation.

                          My 2 cents,
                           Masa K Maeda








                          [Non-text portions of this message have been removed]
                        • Tim Ottinger
                          ... - Refactoring code - improvement of performance - help on other stories - improve test coverage ... aw, heck:
                          Message 12 of 27 , Mar 6, 2009
                          • 0 Attachment
                            ----- Original Message ----
                            > From: Steven Gordon <sgordonphd@...>
                            > Yes, there may be no perfect solution. So, perhaps a hybrid model might work:
                            >
                            > 1. Have the stories driven vertically by verticals teams.
                            > 2. Maintain additional small teams of infrastructural component
                            > specialists (guardians).
                            > 3. For each task of a vertically driven story that involves extending
                            > or modifying a specific infrastructural component, a member of the
                            > vertical team pairs with a member of the guardian team for that
                            > component.
                            >
                            > When component team members are idle, they could do one of the
                            > following depending on priorities:
                            > - refactoring of their component's code,
                            > - improvement of the performance of their component (without changing
                            > functionality),
                            > - help on any vertical stories outside of their component.
                            >
                            > Steve


                            - Refactoring code
                            - improvement of performance
                            - help on other stories
                            - improve test coverage

                            ... aw, heck: http://agileinaflash.blogspot.com/2009/02/what-to-do-when-not-pairing.html


                            Tim Ottinger
                            http://agileinaflash.blogspot.com/
                            http://agileotter.blogspot.com/
                          Your message has been successfully submitted and would be delivered to recipients shortly.