Offending key in /usr/NX/home/nx/.ssh/known_hosts

After upgrading the SSH package due to security issue in Debian, I noticed that I cannot connect back to my machine through NX Client.

The error message is an offending key in the known_hosts. I suspect it is due to a change of the key and thus, when I connect back to it, NX thought the server is compromised.

By deleting the entries in this file resolved the issue.

Finding the right user to test your software

Over the years of development on my product, I have been working with different batches of tester. And I have noticed some good time, as well as some bad time with some of them. I noticed a "trend" that lead to my different emotions.

I feel fruitful and respected when I work with tester who is knowledgeable. The "knowledge" I am referring to, is the understanding of the domain that the software is dealing with. One analogy I can thought of, is the necessity to recruit driver in testing out a car. You won't find someone who don't know how to drive a car for testing.

OK, I hear your voice. You say the test is biased if the user is already "contaminated".

I would agree if the test is focused on the perception on something intangible. E.g. the color should be used or where the submit button should be placed. But if the test require the knowing of the mechanism, it falls into another dimension.

What's more, if your test is composed by a league of such user, I could not imagine how diversify the feedback will be. Each of them will likely to represent the personal preference of the user and what matter worst is, these user will likely NOT to be the same group of user who gonna pay and use your product.

Developing software is an art, but spotting the right gang of people to give you feedback is even so. Now I will go back to my cave and meditate on the next features to implement.

Case sensitivity in namespace prefix affects XML Signature Operation

We were diagnosing an exception thrown in XML signature validation.

<Cannot check signature - Current Node: [#document: null]>

After digging around, we noticed that the XML in process is referencing an element with the name space value in different letter case.

The name space is declared as follows,

<SOAP-ENV:Envelope xmlns:SOAP-ENV=" ... >

But in the transform, we reference the element like this and turns out to be failed

<ds:Transform Algorithm="http://www.w3.org/...>

  <ds:XPath>
    not(ancestor-or-self::node()[@soap-env:actor=\"urn:oasis:names:tc:ebxml- ...

  </ds:XPath>

</ds:Transform>

Did you spot the difference? Yes, it's the case of "SOAP-ENV" declared in the Envelope to be upper case but the one referenced in the transform is in lower case.

Once it is fixed, all works.

Low end dedicated server vs a powerful VPS

Lately, I am having an "itch" in owning a dedicated server, which enables me to install VMWare Server and run multiple of virtual OS. All to satisfy my control freak deep within.

But most of the subscription plan are expensive. You can find dirt cheap bargain but the "server" they are offering is even slower than the PC I have at work.

I have finally found a reasonably priced plan (around US$79) which offers a Dual Core CPU with 2G RAM and 320G of HDD space, not bad in terms of specification. Right before I click the "order" button, I came across this discussion thread, VPS vs. Dedicated - New Benchmark Results. Reading it through and noticed the availability of a benchmarking scripts.

I downloaded the script and have it run on my 2 VPS account, to my surprise, one of them is having a benchmark result 6 times faster than the other! And I am amazed to notice the result of some VPS plan is very close to the dedicated server which I am interested to order.

After a bit of struggling (on money and necessity of course), I opt for a VPS replacement on my slow account instead.

Below are the benchmark result for your reference.

VPS#1 : 241.8
VPS#2 : 164.7
VPS#3 : 29-30

My PC (Core2Duo E6600, 2G RAM, 250G HDD) has a benchmark result of 423.3

Interested to try it on your machine? Download UnixBench v4.1.0 - WHT Variant now!

The unit testing for the web pages - Twill and Selerium.

Unit testing the web application is not a new problem. But unlike the close counterpart of unit testing in Java, where JUnit is the de facto standard, there isn't a methodology that is as common.

I am trying out Twill and Selerium today, both of them has been around since 2004. The two are both scriptable and differs in two aspect.

1. Javascript support
   
2. Browser integration
   

Twill cannot support javascript, for instance, you cannot script it to enter the date value through a popup calendar.

It works in plain text mode so you cannot test out the cross browser functionality.

You can get more information from their web site

Having said that, it is very lightweight and it can integrate nicely with your CI framework.

Selerium address the opposite aspect, the javascript support enables you to test the functionality of your heavily AJAX web pages. You can even use the fiefox extension to create the testing script.

Can't follow what I am saying? Try it out yourself from this URL.

Selenium Functional Test Runner

Oreilly has published an ebook of the Short Cuts series which gives you a quick good read on these tools. Check it out from here