Some of my group mates had taken the course titled "Software Product Line Engineering" from last quarter and came up with the idea of changing the goals of the course to accommodate a software product line engineering approach. Since I was not aware with the principles involved in software product line engineering in the beginning I was skeptical. I insisted that as long as our client (in this case the course instructor) approves it, I did’t see it as a problem.

We were given a go ahead to do the project in a software product line engineering approach. There was not much difference in the end, other than we didn’t focus on implementation of the project at all (as compared to some implementation by other peer groups, all groups in the course were give the same requirements document) and we changed the requirements document to fit the scope of the project (which is kind of cheating, but there was no way around that given the time constraints). We did focus on following the rough guideline of documents (Project Directives, SRS, System Architecture, Test Plan etc). For SRS we took some guide from "Designing Software Product Lines with UML: From Use Cases to Pattern-Based Software Architectures", an interesting book which I would like to read in detail some day. Chapter 14 of this book has a case study which deals with an e-commerce system.

The introduction of SPLE concepts was rather intense considering that not many people in the group had never worked on a document intensive project before. But we did follow a nice Agile approach of managing document (with editing being done on Google docs and final drafts prepared in LyX/LaTeX). The project code was put on google code and it may be released to public (the primary reason for not releasing it is to not hand in a working solution to a group for next year, if the course requirements remain the same). The coding done was basically an implementation of AppFuse tutorial for Spring/Hibernate/Spring MVC (another thing which I plan to follow up in my coming quarter in Software Architecture course)

At the end of the project we had a short verbal discussion with an expert in this field and we came to the conclusion that introducing SPLE in a big bang manner was not a good idea.Another thing which I concluded was that applying software product line engineering in small settings (such as a university) may not be a good idea as there is too much overhead and it’s easy to get lost in abstractions of process. But this second conclusion is based on the experience of this project and may be a better trained small team would perform much better. I would not like to try SPLE unless it’s being done among a team of some what  mature software devs.

At the beginning of each iteration we have a bunch of user stories. Each user story has atleast a description and a set of acceptance tests. These acceptance tests have description and an expected result. Then each of these user stories have a priority for the current iteration. Each user story can be devided into tasks, but this is some thing best done by the programming team. Each User story has an estimate and so does each task. The estimate is usually in number of man hours but it can be less (and it is certainly in our case) In XP since every thing is done in pairs, therefore each task (and probably all with the same user story) go to a particular pair based on bidding (who so ever bids the least time gets the task).

After completion of a task/user story, some one from the team (probably the customer representitive) has to check each user story against it’s acceptance tests.

Each week any two personnel from the teams act as Customer representetive and Tracker for the project. The tracker is suppose to keep track of all the metrics involved in the project. The idea is to make a simple web/desktop based software which can do all of these simple things with no hassle. And since this course is all about agility this software should be up and running over the weekend.

I am thinking of using Ruby on Rails. This would finally be my chance to work with RoR. Perhaps all the facilities are readily provided by other  open source software like Trac, but there are many features like time tracking (like in basecamp) which I would like to see in an Agile Devlopment Tracking software. This is just an idea  for now , lets see if I am able to cook some thing nice

• HowTo: Install Ndiswrapper on CentOS
• CentOS Wireless configuration

I downloaded ndiswrapper source and compiled/installed it. Then i put in the D-Link driver CD-ROM and got hold of Win2k/XP SYS file for the WiFi driver. It loaded itself easlity. I put in the configuration into

/etc/sysconfig/network-scripts/ifcfg-wlan0

and every thing worked smooth untill i got some errors due to differnet channel selection.

The thing is that even though I explicitly say to the server to use channel 9, it some how was using channel 6. Ofcourse I am using the wireless connection in Ad-Hoc more for peer to peer communication. 

Although new Fedora Core comes out in a few days, I will probably stick with Core 6 for the rest of summer. I have decided to stick with Ubuntu on my laptop for the followin reasons.

• What ever you do , there are still issues with products on fedora , which get easily installed on red hat entrprise linux. fedoa != red hat.
• Ubuntu has become a ditro of choice at many places including Google. It is also being preinstalled on Dell laptops, soon it will be on other laptops
• Ubuntu has server and 64-bit editions of their distribution (I even got 64-bit version CDs shipped this time)
• Last and most important, Ubuntu gets shipped free, no one can beat that.

I ordered a free Solaris Express DVD of solaris from Sun’s website. I was surprised to get it within 2 weeks as normally these things reach pretty late in my locality. I popped it into my Intelx86 machine and fired it up. It booted kind of slow, but that could have been my DVD drive. It was not pleasant to find out that the default X server which the installation DVD loaded was not showing up on my Intel D865 based vga.

I booted again in text mode and started the installation process. After a reboot I looked into how to configure X and network. I was unable to get both working in the first try. I went to Sun’s support chat site and I was mentioned the hardwae compatibility link. I tried the Solaris device detection java webstart applet. Every thing checked out (except for my bt878 based tv tuner card). The JNLP mentioned a link for third party network drivers, which I later downloaded (but was only able to get working after some beating around the bush)

Then I decided to visit #solaris on irc.freenode.net (from my other machine)and boy was that an eye opener. First this I learnt is that Solaris is not Linux ( I knew that, but if you even mention name of Linux in #solaris you get flamed easily), second #solaris has lots of rude shrews. They help less and try to establish that their  "brain size" is more than yours since you asked a question. Although looking up on google before you ask a question is always recomended, but it is possible to not find answer to a simple question from a Google search. Any ways I decided NEVER to goto #solaris. I would like to leave the self contained so called UNIX admins to their own sad lonely existance. I went to #opensolaris where people are friendly and enthusiastic (like any open source project).There I got some good pointers.

Solaris has two kinds of X servers;Xsun and Xorg. Xsun is probably a good choice for Sun machines, but Xorg produced better results on x86 machines. The problem I encountered was that the default configuration produced by X -configure was only good for higher resolutions. A utility called kdmconfig lets superuser choose between the two servers. The configuration of Xorg is similar by the use of ‘xorgconfig’.

Solaris x86 express edition comes with alot of goodies which include StarOffice 8, Netbeans 5.5, Sun Studio Enterprise, Gimp, Firefox to name a few. It has two desktops CDE and Gnome. Gnome based desktop looks quite polished (just like Ubuntu).

Overall my first impessions of solris are quite positive. But these impressions are superficial and very cosmetic. In reality I am more interested in knowing more about the internals of Solaris (Kernel and TCP/IP stack and things like DTrace).

Next step was to make collection of statistics from OpenNMS. OpenNMS lacks polished documentation and things are rather scattered. After some playing around with the configurations I have been able to make OpenNMS work on my local network. There are still issues some issues involved namely the addition of an Asterisk category to the front page, auto discovery of SIP and monitoring of  statistics  which are not provided but have to be computed (ie. no of bridged SIP channels on any given system).

I will hopefully post a detailed HOW-TO on gentoo-wiki very soon.