380 likes | 487 Views
Loosely Coupled Sakai. Ray Davis University of California, Berkeley. Deliver usable useful applications in a timely fashion. Goal:. Is it useful? Are you sure that it’s useful?. Method: Empirical. User(-representative) driven Incremental Cyclical Opportunistic refactoring
E N D
Loosely Coupled Sakai Ray Davis University of California, Berkeley
Deliver usable useful applications in a timely fashion. Goal:
Is it useful? Are you sure that it’s useful? Method: Empirical
User(-representative) driven Incremental Cyclical Opportunistic refactoring Loose coupling to framework & services Evidence-Based Programming
Naïve efficiency: Change vendor code directly. Can’t upgrade. Need to maintain unfamiliar code. Loosely coupled: Centralize dependencies. Local implementation. Loose Coupling ≠ Less Integrated
Project management = Risk management Cross-project dependency = Risk Why Loose Coupling?
Separation of concerns Centralization of concerns Avoid redundancy Avoid disruption Improve maintainability Improve testability Loose coupling = standard design principles at a project level
Unrealistic goals Inaccurate estimates Slow refactoring “Living fossils” Unpredictable disruptions “Vendor lock-in” De facto forking Results of tight coupling
Don’t make trouble for yourself. Facades to external services Don’t make trouble for other people. Service APIs What to don’t?
Multiple Moving Targets **SCREEEEEEE…**
Multiple Moving Targets Facades
Application-tailored interfaces to complex or unstable services Minimize maintenance costs Maximize pluggability Reduce costs of unit & application testing Self-document integration requirements Provide fallbacks Facades
Unit tests can be overdone. Generalization for re-use is usually premature. Loose coupling is a leading cause of tight coupling. No Sure Things
Is the framework changing? Will standalone implementations help development & testing? Will implementations be much work? When are facades useful?
Is the framework changing? Will standalone implementations help development & testing? Will implementations be much work? When are facades useful?
Facades Sakai 2.0 APIs Sakai 2.1+ APIs Gradebook GB Facades Tests Standalone
Authentication Context Authorization User Directory Gradebook Facades
public interface Authn { /** * @return an ID uniquely identifying the currently * authenticated user in a site, or null if the user * has not been authenticated. */ public String getUserUid(); Authentication – Who Is This?
public interface ContextManagement { /** * @param request * the javax.servlet.http.HttpServletRequest or * javax.portlet.PortletRequest from which to determine the * current gradebook. Since they don't share an interface, * a generic object is passed. * * @return * the UID of the currently selected gradebook, or null if the * context manager cannot determine a selected gradebook */ public String getGradebookUid(Object request); Context – Where Am I?
public interface Authz { public boolean isUserAbleToGrade(String gradebookUid); public boolean isUserAbleToGradeAll(String gradebookUid); public boolean isUserAbleToGradeSection(String sectionUid); public boolean isUserAbleToEditAssessments(String gradebookUid); public boolean isUserAbleToViewOwnGrades(String gradebookUid); … Authorization – Think pragmatically
public class AuthzSakai2Impl extends AuthzSectionsImpl implements Authz { public static final String PERMISSION_GRADE_ALL = "gradebook.gradeAll", PERMISSION_GRADE_SECTION = "gradebook.gradeSection", PERMISSION_EDIT_ASSIGNMENTS = "gradebook.editAssignments", PERMISSION_VIEW_OWN_GRADES = "gradebook.viewOwnGrades"; /** * Perform authorization-specific framework initializations for the Gradebook. */ public void init() { FunctionManager.registerFunction(PERMISSION_GRADE_ALL); FunctionManager.registerFunction(PERMISSION_GRADE_SECTION); FunctionManager.registerFunction(PERMISSION_EDIT_ASSIGNMENTS); FunctionManager.registerFunction(PERMISSION_VIEW_OWN_GRADES); } public boolean isUserAbleToGrade(String gradebookUid) { return (hasPermission(gradebookUid, PERMISSION_GRADE_ALL) || hasPermission(gradebookUid, PERMISSION_GRADE_SECTION)); } public boolean isUserAbleToGradeSection(String sectionUid) { return getSectionAwareness().isSectionMemberInRole(sectionUid, getAuthn().getUserUid(), Role.TA); } …
As consumer of services Spring-injected facades As producer of services? Loose Coupling
Application = Tool + Component? External Apps App Presentation App Business Logic
Customers End user ≠ Programmer Goals Browser-based workflow ≠ Efficient integration Contracts Functional specification ≠ API Project lifecycles Rapid change ≠ Negotiated stability Application ≠ Service
Project = Application + Service External Apps Application Service Shared Logic
Erich Gamma (Eclipse; Gang of Four): “You can go and expose everything, and people can change anything. The problems start when the next version comes along. If you have exposed everything, you cannot change anything or you break all your clients. APIs don't just happen; they are a big investment.... I really like flexibility that's requirement driven. That's also what we do in Eclipse. When it comes to exposing more API, we do that on demand. We expose API gradually.... So I really think about it in smaller steps, we do not want to commit to an API before its time.” Application ≠ Service
(left as an exercise for the reader) Service change: Implementation
Project = Application + Service External Apps Application Service Shared Logic
From Seth Theriault's Sakai Developer Statistics: 766 lines - GradebookManagerHibernateImpl.java 728 lines - GradebookServiceHibernateImpl.java 252 lines - BaseHibernateManager.java Application logic ≠ Service Logic
Application Service Think Globally:Program Locally Shared Logic Facades to external services