Jacek Wojciechowski's Forum Nokia Blog

Are mobile browsers ready for full web?

jack44 — Fri, 05 Sep 2008 13:25:30 +0300

Looking from developer's perspective, developing for mobiles is not always that easy. If you're mobile web developer, you always have to test your work on many mobile browsers. The Mobile Web Test Suites Working Group has released a set of tests that assess the compatibility of mobile browsers with well-established, current and brand new Web technologies. The test (in the same spirit as the ACID tests) combines in a single page tests for 12 Web technologies. The latest version (1.41) has some improvements over April release. The source code is also available, so one can check how everything is tested.

And the results... are really different for every mobile browser - this applies also to mobile browsers based on "the same" WebKit open source browser engine (AFAIK, WebKit engine powers S60 3rd+, S40 6th+, Android platform and Apple devices, so it can be said that it dominates the 'mobile web') - see the gallery of screenshots for the first version of Web Compatibility Test for Mobile Browsers and the latest version. No mobile browser has passed all the tests so far, but I hope this will change in the near future.

The results for S60 SDK browsers (WebKit based):

As can be seen, there's still room for improvements in some areas (tests: 6, 7, 9, 12, 13, 14, 16). Will upcoming S60 5th edition be even better?

The test is available at http://dev.w3.org/2008/mobile-test/test.html
QR code for quick access:

MS Windows and MAC (unofficially) on N810

jack44 — Thu, 24 Jul 2008 13:11:57 +0300

The number of OSes that can be run on N810 Internet Tablets just grows and grows. Now it's time for MS Windows family. I hope you still remeber Windows 3.1. The trick to run it on NIT is to use DOSBOX utility - x86 PC emulator (the instructions may be found e.g. here or on Internet Tablet Talk forums). What's more interesting, there's a proof that you may also run Windows 95 this way, but don't expect that it will start and work smoothly (you really need to be patient). Thanks to very active Internet Tablet community, you can also have MAC on MIT by using the Basilisk II Macintosh emulator (list of Macintosh Software that can be run on a NIT and instructions).

Purists would extend the list of working OSes by adding Android, Ubuntu and numerous emulators (all ported to maemo) like Amiga 68k - UAE, Apple II - Winapple, Commodore 64 - Frodo, Palm OS - Access GarnetVM, TI-89 Calculator - TiEmu 2 (just to mention the most interesting ones). Can you do this all on iPhone?

So now, the "open platform" slogan gets completely new dimension...

Android on Nokia N810

jack44 — Fri, 11 Apr 2008 13:53:34 +0300

Nokia Internet Tablets are really versatile devices - it's now also possible to successfully install Google Android on them (and it works, of course). The following links describe how to do it well: check here and here. Have fun with comparing maemo platform with Android on real device...

S60 and creating applications with quality in mind (part 3)

jack44 — Tue, 10 Jul 2007 13:55:17 +0300

This time I'll continue the subject of QA (part 1, part 2).

As you know, probably the only way of discovering all defects in a product is long-term use by many users. Unfortunately this can be achieved after the product is on the market. Since this kind of feedback would be very valuable, user experience should be simulated in some way. TRUE should help to solve this problem. According to "S60 Smartphone Quality Assurance" book, Testing Real User Experience (TRUE) is normally carried out as soon as the product can be used in a meaningful manner, in other words, once the product has enough functionality to be used by anyone. It provides feedback from real-life usage during the R&D phase, when fixing is still possible. TRUE testing doesn't need any test cases because a selected set of people from a target consumer group are using the product prototype in the way they want to. These people are trained to report all defects they discover in the product. So now everybody can understand why some people from sites like allaboutsymbian, my-symbian or others reports that their 'protos' works slowly (in prototypes many things are logged to provide some valuable information).

Please note one more thing - it has also been said that "one fix creates five new defects in the system" (sad but true).

S60 and creating applications with quality in mind (part 2)

jack44 — Fri, 29 Jun 2007 14:24:10 +0300

Last time I wrote about EDFU, but this time I'll focus on some testing techniques. Probably many of us use them, but it may be interesting to recap.

We can use:

data-driven testing - to find defects in which certain data is incorrectly processed (e.g. using values above a higher boundary or below a lower boundary can be very effective in finding defects)
logic-driven - to identify all incorrect handling of the logic (e.g. testing certain combinations of inputs to every logical expression)
event-driven - to discover incorrect handling of events (the time sequence and arrival time of different inputs may introduce failures, especially in 'mobile world')
state-driven - to find all incorrect state transitions (all critical states and critical transitions need to be tested)
dataflow-driven - to focus on problems in component interfaces (data element receives a value in one place and uses it in another - sometimes the value is misinterpreted)

One can also add Message Sequence Charts to understand client-server architecture better, code review (probably one of the most efficient defect finding techniques, but it's very slow and should be used only on the most risky components), extreme programming (includes testing).

Some other facts related to development and quality assurance process you can find in a new set of S60 guidelines: http://www.forum.nokia.com/document/S60_Platform_Development_and_QA_Process_Guideline/

I'd like to know your opinions and experiences in using testing frameworks to speed up testing like JUnit for java (or checking your code with SymScan for C++)... Do you consider the Symbian Signed and Java Verified requirements when designing test cases?

S60 and creating applications with quality in mind

jack44 — Wed, 20 Jun 2007 21:24:33 +0300

I've just finished reading "S60 Smartphone Quality Assurance" book, written by Saila Laitinen. This led me to some thoughts about quality of final products. One part of the process of creating the product is testing.

Probably many of you are familiar with words like 'constructive / destructive testing' or 'white-box / black-box testing'. But it may be challenging task to decide which part of the code to inspect the most. One way is to use Estimated Degree of Functional Usage (EDFU). As Saila wrote in her book "EDFU is a numeric value that provides a simple estimate of how often the end-user uses a certain piece of functionality. To determine this one needs to understand consumers' behaviour very well. If the value is one, it is estimated that the end user uses this particular piece of functionality every time he or she uses the product. If the value is close to zero (e.g. 0,01), it is very unlikely that na average user ever uses such functionality." A good question to ask is how to define an average user. Normally, we target our applications to certain customer groups. When we find people that match our profile we can ask them to use our prototype application and watch how this potential customer group uses it. "The higher the EDFU, the more important it is for that part or function to work properly. However, even such features that have a value 1 for EDFU are not automatically equally important from testing viewport. Those features whose implementation results in architecturally complex solutions should receive the most testing attention." Of course, this idea can be used in the process of developing not only mobile applications.

What do you think about EDFU? Would it be useful in your cases?