Thursday, January 31, 2008

Technical: MIDlet Session Handling

Found some example codes on Session Handling for MIDlet. 

Invoke the servlet...



Prints out the Session ID


This is rather last min and we have yet to fully understand the coding, thus we are not able to demo this portion in the end of Session 1 Presentation. 

Wednesday, January 30, 2008

Technical: JSP and Servlet Communication

The diagram below shows the best practices of JSP and Servlet communication implementation.

Tuesday, January 29, 2008

Updates: Captha Design v3.2

Symbols were added in to further increase the sample space. As we know certain symbols like "," and "." may look alike from certain angles/from additional background noises, we decided to choose 10 more obvious symbols to start up with. They are  { '!' , '@' , '#' , '$' , '%' , '^' , '&' , '*' , '=' , '?' }

Sunday, January 27, 2008

Technical: Captcha Breakdown Structure

Below is the detailed breaked down of the captcha structure.

Technical: MIDlet JDBC & mySQL

The CLDC framework do not support JDBC APIS. To connect to databases we can call a Servlet from the MIDLets. Connection to database, running the queries and other functionalities have to be be implemented in the Servlet. The MIDlet passes the parameters like username, password in a string throught a DataOutputStream to servlet

Tuesday, January 22, 2008

Updates: Captcha Design v3.1

To tackle the bad design of the previous version, modifications are done by replacing the usage of random range of angles of values -45 to 45 degrees to randomly select 4 fixed different angles degrees. 


At least now we wouldn't have to worry about characters going upsize down/off sight.

Sunday, January 20, 2008

Technical: Prototype using Texture

Based on the "blending text into the background" concept, we created several prototypes of the CAPTCHA image. Following the techniques highlighted in lex's post on Java3D Appearance/Texture,  we alter the appearance of the 3DText with the mapping of a texture. Using the same texture for both the TEXT and the BACKGROUND gives us the following results.








To conclude, it is either our mapping technique s**ks or ... considering the poor results and also justification that texture takes up space (may not be feasible for mobile phones) .. we decided to drop the idea for the time being.