Retrospective

I was going through Slashdot and saw an article (http://ask.slashdot.org/article.pl?sid=09/02/28/037256). It starts with the author saying that He's been using GNU-Linux for the last 10 years. I realized that I have been using it for more than that... about 13 years now!

I've seen it come up from a difficult to install, geeky (the list goes on) to what it is today; a competitive desktop OS! I've gone through so many distros, spoilt so many working systems and played around with some FOSS. I got particularly interested in GAIM (now Pidgin) through a senior and friend at college.

So where am I finally when it comes to it. I'm a big supporter and consumer of FOSS, but my contribution has been minimal. Recently I reported a couple of bugs in the Geos library while building it on the Sun Studio tool chain. So basically that's the nature of the contribution I've made.

I do feel that the amount I'm taking from this unlimited pool, can not match the few drops I'm putting in. But I guess a lot of people are pouring in buckets one after another. That's what allows people like me to get away and still be at peace.

There was a poem I composed long time back... I'll post it some other time. It ended something like this:

Then a flash ripped me apart,
As I tried to run,
With guilt still within my heart...

Lets see how far the Icarus flies!

Way ahead

So I've finally decided what I'm going to do without looking back. I'm going to apply for MS/PhD for next year fall. I've started research work I want to get out before I apply. Its basically in the two areas I have been working most: Geo-coding addresses and Path finding.

Geo-coding for a country like India is beyond the realms of mathematical complexities. Yet, its something that must be done. I'm taking a different approach this time around. I'm working on getting parameters based on both textual match and geographic proximity to find a match. This approach, I believe, can work for a much wider array of input types.

The other, path finding, is a problem that has been worked on a lot by a lot of well renowned people. My idea is to develop an optimal parallel algorithm for a transportation (road) network.

Good luck to me!!!

Bad Times

I was chatting with a friend from college who works in a mid-sized product development company. In fact, there was a time when I was thinking of joining the same. So she tells me that the CEO made an appearance in a Maruti 800 instead of his Accord and announced that they were in a really bad financial position. Basically they may not be able to pay the employees.

About time I got more serious about what I want to do! Always been in this dilemma, and it keeps coming back. Go ahead with academics and research or stay in the corporate world. I am and know can continue to do well in the corporate world. But not sure whether that's what I want. At times it seems like bliss. But, isn't ignorance bliss?

Guess I'll finally put some serious thought into it and make the final call... Till then, ciao!

Trying to clean up

I wrote about the mess I had gotten myself into. Almost out of it now. But one big learning out of this is however big a brand may be, go ahead only after root level testing.

Thought of a couple of ways to make the direction search faster and use lesser memory. Going with the good old C file approach. Think it will work. On paper looks great... Reducing the graph size by about 6 times. Speed figures I will get once I analyze the algorithm. Got to look for a parallel algorithm for routing. I was thinking that if a PND with a 500 Mhz processor was able to calculate routes so fast, I should be able to do much better. Lets see how it goes!

In a mess

The first time I took such a bad decision... I convinced the top to buy me a couple of Sun UltraSparc T5120 machines for our maps portal . All was well and rosy. I couldn't stop blabbering about these monstrous machines. Post setup and deployment, it goes into the testing stage and WHOA! Search times have gone low by at-least 10 times!

Traced it down to the fact that Lucene (one of the components I use in the search engine) does not run well (?) on UltraSparc platform! I split the index and used the ParallelSearcher on it. Result was still the same. No idea how I'm gonna go about it. For now, I've put this part on an AMD Opteron machine and its doing well. But I have no answer, except a sorry to give to my management. In times like this, a mistake as big as this might just cost me a lot more than it would have otherwise.