Stand on Others Shoulders Knowledge Sharing

1
Stand on Others Shoulders! Knowledge Sharing
Software Development

• N. Rammohan

2
Introduction

• For efficient, fast, and high quality software
  development
• Communication and diffusion of information is
  very important
• Effective experience sharing
• Current software development model is distributed
  (because of outsourcing)
• So increased importance to above factors

3
Need for it- example1

• Task- test the new s/w build on h/w platform
• Known general procedure
• Step-1 Load the build onto the h/w
• Step-2 Load some configuration file
• Step-3 Run and test for minimum requirements
• But for a certain base build, certain h/w, it was
  not working!!
• Tested in all known ways
• Took more than two weeks because
• Problem- Just run extra script after step-2
• It takes just one second!!
• It happened in Qualcomm- a big MNC
• What about open-source communities?

4
Example-2

• Working with latest code uploaded to server
• Uses some new database engines
• Have run into unsuitable driver problems
• Tried in all possible known ways
• After lot of effort, the issue was found
• a wrong JDK version was used!!
• This kind of experience can be captured
  automatically and re-used elsewhere
• When the same code and JDK version is used

5
Contd

• Interdependence of all s/w components
• Difficult to isolate the source of error-
  experience matters
• To solve a problem
• Users 99 of the times try to do permutations of
  all things they know
• Trial-and-run experiments
• But there may be something that somebody knows
  which can solve the problem instantly
• Benefit from reusing ve or ve exps of others
• Management of who knows what knowledge is very
  critical in corporate s/w development projects

6
Contd

• But users may not be willing to document their
  experiences
• Automatically capture and disseminate such
  knowledge
• Two kinds of knowledge
• Explicit knowledge, which exists in the form of
  formal artifacts such as documents or source
  code.
• Currently, they are buried in private workspaces
  and not available for interested peers.
• Implicit knowledge, that gets seldom formalized
  such as how to resolve a certain error.
• Part of this knowledge can be captured
  automatically (such as an error and subsequent
  steps that try to resolve it).
• There should be a way for the users to document
  them with minimum effort
• Current IDEs provide only explicit knowledge
• Ex. Resolving syntactical ambiguities in code
• Intelligent IDEs are needed to support
• Advanced error handling
• reuse and collaboration

7
Goals

• A plug-in into the Integrated Development
  Environment
• Monitors users all coding and development
  activities
• Retrieves and stores knowledge
• Gives recommendations based on past users and
  OTHERS knowledge
• Makes software dev more robust, fun, powerful
• Ideally you never repeat the problems faced by
  others!!
• You are not alone in coding- watched by all peers
  through their knowledge bases
• Intelligent IDE- resolves some non-trivial
  semantic errors automatically

8
Existing solutions

• Centralized units for organizational learning
• Proved to be useful for large s/w companies only
• Not optimal for open source dev, distributed dev
• It eliminates contextual aspects of knowledge and
  creates a general representation and provides a
  knowledge package
• Contextual aspects are very important to capture
  users personal and working contexts
• Knowledge which is not useful to one (e.g., for
  an expert) may be useful for others (e.g., for a
  novice)
• Universal representation of knowledge is very
  difficult

9
Existing solutions (Contd)

• Lightweight knowledge storing approaches
• Often NOT integrated into IDE
• Based on Community of Practice (CoP)
• Social learning among people with common interest
  in some subject or a problem
• Ex. Weblogs, wikis
• Issues
• Wiki becomes increasingly useless overtime
• They collect more and more unstructured knowledge
  without a chance to find anything useful
• An appropriate usage methodology and powerful
  search mechanism are needed

10
What a good solution should have?

• Represent knowledge in highly reusable form
• Knowledge relevant for s/w dev is hard to be
  formalized w.r.t its efficient reuse
• Should consider knowledge levels of different
  users
• Info relevant for one could be very relevant for
  others because of different exp levels
• Contextual info is very important
• Should consider distributed s/w dev model
• Very imp because of increased outsourcing and
  open-source dev model
• Consider heterogeneity of used vocabularies
• Should include personalized proactive recommender
• Context-sensitive semantic search

11
How?

• The semantic technologies
• Associations between meanings (interpretations)
  and data, to be known and processed at run time
• Improved search
• Discovery and content provisioning
• Dynamic integration across distributed systems
• Used in making Semantic Web
• A machine understandable web with smart data
• Context observer- extracts the working context of
  the user

12
Contd

• Semantic recommendation
• Delivers knowledge items based on actual working
  and personal context of the user
• Assume purely distributed user community
• So p2p infrastructure for decentralized
  communication among local knowledge systems
• When to search local databases and when to extend
  it to the global network is an issue
• And finally the Knowledge Desktop
• Users interface to the system

13
Tightening knowledge sharing in distributed
software communities by applying
semantic technologies
14
Players
15
?

• Questions are always guaranteed but not the
  answers!

16
A prepackaged knowledge system with centralized
storage- example
17
What we aim for?