Wednesday, March 16, 2005

Disposition

A gave me some input on how to disposition the thesis

State of art. Describe the existing systems. Theory.

I've found an increasing amount of WCMS (as you can see on my links), and a lot of these are based on open source, particularly php, but I believe java will soon catch up (http://theserverside.com/news/thread.tss?thread_id=33163).

Also note in the discussion between KM and CM, the CMS fall into categories like Groupware, e-Learning, Portals and Wiki. These are classical KM IT-tools, but still the KM protectors have a belief that CM is nothing but the treatment of data and information. Au contraire (*sp), but we've found our own word for that: Content Repository, and about everyone who are dealing with CMS these days are very excited about the coming Content Repository standards and implementations, either you're talking about WedDAV or Jackrabbit.

Also bring in discussion on the work of others on the subject. Write down summaries of the articles and books I read.

The requirements, the users. How we gather requirements for the WCMS.


In P, we recently had a customer session digging out the requirements from our users. We're applying a lightwheight form of the Scrum methodology, which is a lightweight variant of XP, which is again supposed to be one of the lighter, big-M Methodologies.

The point being is that we model/assign requirements as "User Stories" (written by the customer, max 3 sentences, describes a function in the system) instead of classical Use Cases. We scrambled together about 50 user stories, most of which can be applied by all our different customers, the rest are very customer specific.

Example of customer specific functionality is NAAF, where a professor up in Trondheim feeds pollen-data into the CMS so that is published NAAF's webpages (in a much more readable view), and syndicated to a set of subscribers through email.

The User Stories will lead to Acceptance Tests, which will together with software architecture make up the Unit Tests. Also bring in discussion on the work of others on the subject. Write down summaries of the articles and books I read.When we have solved the Unit Tests, hopefully the Acceptance Tests will also run successfully, and then the product will be complete. Unfortunately, by that time, 10 new User Stories have emerged. The key here is to have a component architecture which makes it easy to integrate new functionality as components.

Prototype. Design.
Appfuse, Jackrabbit, content repository JSR-170, portlets JSR-168, JSF and the middle-tier made flesh&bone oughta do it!

Use of CMS. Validate the system. Metrics.
Now this is trickier biz. As we are an external organization we have no say in how our customers measure the inward improvement in workflow and document process. We should however supply the customers with some tools and methods for doing this. Perhaps some surveys now on how publishing and web editing is done, and how much manpower is spent.

Thoughts about WCMS

Originally, my problem defintion was something like:

The development and analysis of a knowledge management system.

Now it turns out that what we are producing at P is purerly a web content management system (WCMS). This might sound rather dull and unsophisticated, but it is important to note that "P Portal" is a general, but customizable WCMS.

I will try to elaborate on these two terms:

A WCMS is often developed inside the organization for use in that organization only, and it satisfies the organizations need for web content. An example is a small business hiring a web-designer that makes a small website, where news can be added on the front page with the help of a simple form. Another example is a larger corporation hiring in CMS consultants to produce a WCMS which is seamed into the existing CMS (also known as ECM-system).

A general WCMS is what is often referred to as a (W)CMS package, something you buy, install and start using the way it is off the self - out of the box.

A general WCMS is not specially developed for an organization to suite that particular organization's needs. A general WCMS has the ability to fit into a set of organizations. The size of this set depends on the customizability and plugability of the WCMS, as well as the organization's requirements.

It is important to note, as corporate websites are becoming more and more critical, that these requirements are increasing. Design, functionality and content are typical features that attract customers. A company can often be judged by its website, altough I will leave out this discussion to marketing theorists and analyzers. The result is that organizations demand WCMS that can produce stylish websites with new functionalities like commenting, news-feeds and online ordering. These modern functionalities are typical aspects of traditional CMS (collaboration, cross-media syndication and process-management, respectfully).

Following this trend, the amount of potential organization that can use one general WCMS rapidly decreases. You simply can not develop one system that will fit in to a larger set of organizations because they want different site-design and functionality.

This is a challenge that P Portal is facing.

There is a pretty straight forward solution to the problem: Make a WCMS that takes all possible requirements into consideration.

This is of course impossible, but what we can do is to:

1. Define the functionalities in an abstract way, so they can be implemented to satisfy a wider set of requirements..
2. Make sure the functionality is extendable to future requirements.

I like to divide requirements into visual and functional.

Many WCMS are turned down because they do not look the way the web-designers want. A WCMS is inclined to standardise content, and this reduced the variety and visual experience of a website.

Following our two guidelines above, it would be natural to seperate the visual face of a website from its content. Seperating view from content is an ancient but still healthy CMS paradigm.


So the webdesigner has to go to work on something which is not content specific, typically several template pages that are used to render content. Depending on the skill of the webdesigner, she enjoys going to work with a wysiwyg/drag'n'drop editor (like MS Frontpage), pure html, css, javascript, flash and so on. Some of these formats do not mix well with middleware-produced content (typically XML). The details on how this is done varies immensly by choice of platform, but sooner or later, the webdesigner's template has to interface against the content produced from middleware.

If the designer has a proper and elegant middleware interface, it is easy and straightforward to insert content into the template page. Evidence of the opposite is the mass of ASP, JSP and PHP pages around on the net that contain programmatical logic, and the HTML is so littered with code-snippets, scriplets and scripts that webdesigner, much less the wysiwyg editor can read the page and make it look the way they want.

There are frameworks that supply interfaces like this. I am (un)fortunately only conversed with the ones that are used in JSP. These frameworks clean up the HTML by keeping programmatical logic inside custom made html-tags. Struts, JSTL and JSF are examples, and it is of course possible to develop custom tags on your own.

With a solid and sufficient interface, a webdesigner can surround the content with as flashy and fancy view as if she was designing a simple HTML page. There are other issues regarding the view of the site, particularly how navigation is handled. Imagine the menu bar on the left or top of a front page. It often reflects how the content is structured. This is mixing of content and view, and should be worked around somehow.



The functional requirements are trickier to sort out. One website needs a forum, another needs a webshop and article-publishing. As stated above, we need a suite of tools that is both (1) abstract and (2) extendable.

A good starting point for predicting future requirements is CMS theory. CMS has been around for a long time, and it is not often websites are imposed to support a feature which was not allready supported in the organization's CMS. Of course it does happen, the weblog being the classic exception of the rule.

CMS theory produces the following typical features:
-Document management (publishing news, papers) - primary input
-Digital asset management (file repository)
-Feedback (comment, forum, survey, blog) - secondary input
-Business process management
-Digital rights management

These belong to document management:
-Versioning
-Collaboration
-Internationalizatioe

So how to we abstract the solution of these? Stay focused, merge similar documents, keep it generic and dynamic. How do we keep the solution extendable? Guess we have to revert to the architecture of the software here...

//TODO finish this up



First task definition

Task definition for T

Development and analysis of a Knowledge Management System.

The case of research is the software produced by a company called P.

The thesis will consist of an analysis and development of a knowledge management system, P Portal . The solution should be possible to re-use in several different organisations.

P Portal is a publication tool developed in Java 2 Enterprise Edition through design and programming of Servlets and JSP-pages. The solution should be module-based around a central core so that different modules can be added at choice to expand the functionality of the portal.

The thesis will describe the process of development of the portal and the different modules. Documentation will be included on how the solution effects the targeted organisations as a knowledge management system.

P is an established company with several large customers using version 2.0 of it's software for their online portals. T's task wil consist of developing new and improved functionality for the 3.0 version.

The spesific tasks in P:

- Developing modules for
- User management
- Dynamic menus
- Editing and publishing
- File archives
- Picture archives
- Forum (bulletin board)
- Search engine
- News mail
- Surveys
- News agent
- Communication (back end to front end)
- Internationalization
- Analysis
- Documenting the development process and the system

Milestones:

  • 31st of March, 2005 - Alpha version (core functionality)
  • 31st of May, 2005 - Done testing Alpha version
  • 30th of November, 2005 - Analysis and development of extension modules
  • 31st of June, 2006 - Done writing thesis

The process and progress of the thesis will be overseeen by A.

S will be guiding T's internal development of the solution.

Aliases used in this blog

To anonymize and ease the content of this blog, following abbreviations are used:

T - Me (Thomas Ferris Nicolaisen, no secret :) )
A - My master advisor
P - The company where my master thesis has been externalized
S - The external advisor at the company
U - The university I'm studying at