Badri

a brave new year

badrix — Fri, 08 Jan 2010 01:07:52 +0000

Here are some awful truths I’ve realized in 2009.
1. New year resolutions don’t work, but still I get tempted to sneak in a couple of them every year.
2. There is no rule which states that you have to keep blogging once you get started.
3. jquery is the best thing since salted cashews.
4. 93% of my peers haven’t heard about Joel Spolsky.
5. Nobody avails the “help” menu in a program, let alone read it.
6. Wii is better than the other 2 consoles out there.
7. Twitter is a site where a bunch of people write in 140 characters about what they’re doing, and it is a smash it.
8. wikipedia still does not display ads.
9. India actually has 3G.
10. vi is good, emacs is wow!
11. python 3 is an indication of how reluctant people are to change.
12. Atwood’s law is slowly becoming a reality.
13. If you put “green” before any concept, it has greater chances of selling.
14. Copenhagen accord was a colossal waste of time and money.
15. Steve Yegge is a real person. I even stumbled upon a photo of him.
16. Java still seems weird, and I hope scala takes over the world by a storm.

updates

badrix — Sat, 29 Aug 2009 02:36:17 +0000

A word of note for the few souls reading this blog. I’ve started a new one. Its more about the things I tinker with on weekends and late evenings. Warning: Not for the technically challenged. I’ll still continue to post wierdos in this blog, and try to be a touch more regular in doing so.

top weasel words which actually don’t mean anything in most contexts

badrix — Fri, 12 Jun 2009 01:19:36 +0000

I call them powerpoint words. Feel free to garnish them in your presentation and see your pointy haired boss go wild with joy!

online learning/e-learning
agent of change
object oriented design(subjective)
strategic meet/strategic meeting and any variant involving strategy and people meeting together
collaboration
paradigm
write up
synergy
SWOT
methodology
proactive
key areas
state of the art
mashup
lastly, for best results enclose all your crap in bullet points…

The Big company gets it wrong again

badrix — Tue, 05 May 2009 16:22:55 +0000

Nothing beats plain old paper and pen for discussing ideas and brainstorming. No matter what web technologies you have, its a far cry. I always carry a pen and a notepad, because you never know when an idea might strike you.The other day, for some reason, my notepad was missing. I just wanted to draw the prototype of some design. I asked most of my neighbors for paper, they didn’t have any(symptom #1 for a Big Company, your peers just work like mindless bots, no planning, thinking etc). I reached for the whiteboard. As it turned out, 2 markers were present, both dry of ink. I was aghast by the sorry state of affairs and marched straight to the guy who is in charge of all this stationery stuff. He told me he couldn’t give me any supplies directly, and that I needed to apply via a CRM portal even to get a ball point pen and a post-it note(they were stored in the next room, mind you!).
I asked him as to how to do this CRM thing. He mailed me an excel sheet morphed into a registration form to fill up. I went back to my location(2 floors away) and filled it up promptly. A couple of business days later, I got a mail telling that my request for the CRM portal registration had been approved. Another symptom of Big companies. People work slower than sloths until you push shove their asses to speed things up. That’s basically got to do with the company’s attitude in general. I logged into the portal and made a request for a notepad. Their basic design was, when I try to add a new item to the list, it pops a new window and I am supposed to click the needed items, which get appended to my list. Guess what, the pop-click-add sequence does not work in firefox. Muhahahah! After 2 days and one horrifying user experience later, I am still without a notebook. You could as well ask me why am I not placing the request in internet explorer, alas, I am having a linux distro on my machine.

Some random things which are connected to this incident:
1. nobody in my team knows till today that they could procure stuff through that CRM.
2. I got half a dozen notepads and a couple of markers stashed under my desk, just in case.
3. I started hating big company paradigms like “overdoing small things” and “needless hierarchy” with a passion.
4. Because of overwhelming (poor) response, Big Company is following similar practices in other aspects as well. They call it cost cutting.

how to lose customers effectively

badrix — Fri, 01 May 2009 17:14:01 +0000

These are some anti-case studies in user interface design. Being an Indian IT male, it is not uncommon for me to use the Internet for a lot of sundry activities. This does not include mail checking, twitter and other narcissistic stuff. So, we are effectively left with reading blogs and doing housekeeping jobs, like online banking and bill payments. Online bill payment is cool. But the whole process is ridiculously hectic. At the end of the day, it is a bitter experience. Even if it has saved you a lot of time.
The other day, I had to pay my Airtel bills. I was scrolling the website as to how to go about it. Being a first time user, I wanted to register with the site. Their “register” button was the smallest in the page!
After reaching the registration page with much effort, I had to fill out the usual registration form. For some reason, I had to enter a date. It had plain selector controls for choosing month, date etc. The year column simply did not WORK! Whatever year I selected, it greyed out. Simple as that. The the form validation fails and politely requests you to enter a valid date. I was helpless! In the age of javascript calendar pickers, they had a selector control for dates. I don’t object that, because not all people in the world might have bleeding edge browsers. Failing to check if the form works successfully in common browsers, is unacceptable. They potentially lost a customer because they did not test their site. Not that their service sucks (it’s actually wonderful), but your customers do mind a lot about presentation.
The other day, a similar thing happened with Vodafone. In the first place, Vodafone’s is one of the clumsiest websites I’ve seen in my entire life. Web design courses can take it in their courseware as an apt anti-case study. As if this were not enough, their submission form does not ‘submit’ at all. You click the submit button, and it sits right there without moving a muscle. What the hell is happening? The user has no clue. I visited a Vodafone care person, he told me the server was down since yesterday. Maybe all these guys are trained to give the “server down” thing when something is wrong with the computer, you never know.
We come to the third anti-case study. This is not web centric. It is related to phone banking. When I dial the number of my bank, it gives me an automated voice response dictating all menus and choices they have got. No problem here. I choose some option which prompts me to enter my card number. Nowhere in the whole instruction sequence they mention that I have to enter my 4 digit pin. After reaching the fifth choice in the third submenu, they tell me that I had to enter my pin back in the first menu. By then, 40% of the whole call is spent in traversing menus and another 30% in their advertisements and new offers. No work done effectively. Consider this if you are in an emergency!
I fail to understand why they don’t do hallway usability testing before releasing a product. What’s more, I had complained twice to Airtel, Vodafone and once to the bank folks that there is such a problem with their user interface. At the time of writing this, it has still not changed. Bad user interface is one way to lose customers, not listening to them will actually make you lose them faster.

The day my sound card stood still

badrix — Sun, 30 Nov 2008 07:53:37 +0000

I am one of those dual boot nuts. Let me digress the details. I am not sure whether I will be able to feel at home while editing about 15 files scattered across my screen in the form of emacs windows. I boot linux for that purpose. I have this feel good factor when I am coding in linux, or for that matter, any UNIX variant where you tap your keyboard a lot more than you click your mouse. For in house entertainment, I boot windows. “Entertainment” means anything from playing Beethoven’s Moonlight sonata mp3 to watching The Simpsons complete season 17. Its hard wired in my brain that “linux=programming productivity” and “windows=fun house”.

I tried a sinister experiment then. Why can’t linux be “fun house”? All I needed was a decent movie/music player without all frills. I downloaded mplayer and started playing an mp3, and to my surprise, everything looked ok, except a minor fault. There was no sound!

I strained my ears to listen an familiar moonlight sonata tune, all I could hear was a faint bhool bullayya tune, which was apparently coming from the next door. So, I thought that it was time to configure my sound card. I opened one of those “configure your sound card” programs and fiddled with it. In windows, it would have been a different story. Probably a yellow coloured dog(the one featured in windows desktop search) would ask you smart questions like what is your sound card make, show a magnifying-glass-hovering-around icon to convince the user that its probing the hardware etc. A couple of minutes later, it would play a familiar “ding” and ask if you’re able to hear it. Thats it! Let there be sound. Back to our linux configuring story now….

No yellow dog here, the program looked terra incognita to me. Some of the things I googled for:

configuring sound card in linux
SUSE 10.2 sound card problems
YAST sound card
HP nx6110 sound card
nx6110 sound card problem
HP nx6110 SUSE 10.2

If you are a good observer, you can notice from the search patterns that I might be slowly hitting a dead end. You are damn right. Two hours and 3 coffees later, still no moonlight sonata playing on my desk. Despite so much caffine, I was exhausted. I had charted out 2 ideas here:

code in a quiet environment, that would actually increase my productivity(would it?). I could still watch movies and videos, and my video card was working like a whistle. Maybe erstwhile scifi movies of the silent movie era, like metropolis would be a good bet.
boot windows for doing any multimedia thing connected with sound. That sounded like a sound idea.

At first, it seemed odd. But then I began to realize the subtle fact that how much hearing to music has an effect on productivity. You will have to repair your sound card to experience the results yourself(Don’t try this at home). This applies for windows users only. For linux users, no need to muddle your hardware at all. Just boot a lesser known version of your favorite distribution! It will be quite(quiet?!) an experience. Better results guaranteed if you have older laptops.

These quiet days, I spend a lot of time on coding. Did I mention that I got an i-pod recently? I need not reboot anything. Its just that my thumb strains a bit when I do the Martin shuffle.

I? robot?

badrix — Tue, 11 Nov 2008 08:13:21 +0000

The author believes that computers will pass the Turing test by 2029. But, he muses what will be in store for computers after that.

I never won a game of chess against humans, save a few exceptions like my grandma and 5-year-olds. It made me probe as to what makes computers to win a chess game always, whether its your grandma, or a grandmaster. But not any more. Computer chess is not considered “true” intelligence. Its just about crunching numbers and matching patterns. AI proponents now go in search of newer vistas like making computers play a game of go, compose music etc.

Why is Turing test considered to be the litmus test for computers claiming to be intelligent? Because, language is one of the most complex functions of the human brain. A computer, says Ray Kurzwell, can get away with a few inconsistencies in other departments like writing poetry, composing music and painting. But, does Turing test tell whether a computer is conscious? If so, how?

The better question to ask would be, is consciousness an essential attribute of intelligence? The definition of consciousness is a topic in itself, which has been beaten to pulp. I buy this as a definition of consciousness: “any entity which is aware of its existence is said to be conscious.” We’ll see some grey areas which will lead to more greyer areas.

grey area #1:

Is an ant conscious? –> Is there a test to determine whether an ant is conscious or not?

grey area #2:

There is an all new Acme ultra-scanner device attached to your brain. It understands neural signals by converting them to electrical signals. We can attach it to our heads to digitize all our experiences, tastes, phobias and high school crushes and feed them to a computer. Of course, the computer will be our alter-ego. Will it exhibit consciousness? Will I exist in two physical entities at the same time? sounds like science fiction to me. –> Is consciousness just a chemical reaction in our brain, which makes us self aware time to time? When I am sleeping without having any dreams or nightmares, where am “I” ?

grey area #3:
are we conscious?

Douglas Hofstadter feels that our brains are too weak to understand themselves.I think we are being too anthropomorphic by making computers to become self aware. Are we sure that we are self aware? Do we just assume that we are? Does it vary by degrees? I have my doubts. In this regard, grey area #3 leads us to many interesting dimensions. One of them is the realization that all mammalian feelings are just chemical reactions. For instance,

hunger – triggered by hypothalamus, fluctuation of leptin
thirst – brain signals “thirst” when osmolite concentration is high
humor – signalled by motor nurons in the frontal lobe area
love – dopamine(?)
Godliness – triggered by “god spot” in the brain

Is it God’s way of tricking us into believing that we are self aware? In that case, we are computers again, no better. Just cellular counterparts of today’s Turing test contenders. I believe that evolution is taking subtle twists, that it need not proceed linearly, or even biologically for that matter. Natural selection will give way to artificial intelligence. Its too microscopic a timescale for me to act as an observer and predict silly things.

Hopefully, computers won’t exhibit kleptomania like in the movies. They won’t long for our affection/acceptance(like in the movies too!). If an answer is figured out for the consciousness question, we can peacefully co-exist without any secret “world domination” plans.

what the big company got wrong

badrix — Fri, 08 Aug 2008 03:57:06 +0000

The opinions expressed here are the author’s own. He is not the mouthpiece of some big company.

Back when I joined Big company in the silicon valley of India, I had certain outrageous assumptions about the way a big companies operate. Lately I realized that most of my assumptions are falsified. I wanted to take stock of that stark realization and capture the moment/thoughts. Hence this post.

When you dig deep as to why Big Co fails to perform as good as it used to do as a small co, you will have no surprise in your findings. They make blunders, not mistakes. Why would anybody make blunders, knowing that the consequence is going to be undo-able and the repair costly? Because they do it over a period of years, very slowly, without realizing that they’re erring. Reason #2, if mistakes are distributed across people, nobody owes responsibility, nobody feels guilty and nobody acts. I can’t blame a single person for inflation or price hike. If on the other hand, I feel bad about inflation, and instead of kicking around I start being conservative on resources, start using public transport for commuting to work etc. I act as if I am also contributing to the price hike otherwise(which is sadly true). On to Big Co’s blunders now.

1.substandard products

The most basic thing Big Co. gets wrong is product quality. If you are delivering a product to the next layer(it can be your customer or the quality team), you make sure that you get things done right the first time. At least strive to do so. Big Co. thrives on hope, numbers and hype. It always ends up succumbing to market pressure, hence deadlines dominate all business proceedings. They take this as an excuse and deliver products which actually suck. There is no distinction between production quality code and code written by college freshmen in Big Co., barring a few instances.

2. Premature optimization

Optimization bug bites Big Co’s employees at the wrong time, always. They spend a lot of developer time is squeezing out CPU cycles out of hardly used code. The reason they give: Give users an edge over competitor’s products. Correctness of code is more important than speed.

3. Poor planning

I agree that market decides Big Co’s plans. But the deadlines should be scheduled optimally so that there is a win-win situation for both the developers and customers. Afterall, that’s what managers are paid for. But Big Co. plans dates without even consulting a single developer, and impose it on them. Most deadlines are not enough for delivering a complete product, resulting in delivering a sub-optimal solution which is not fully functional. I call it a half-bake. The reason Big Co. fails in this aspect: The take the “release early, release often” idea too literally. Would you be happy if the the pizza guy gives you a pizza in 3 minutes, provided its just half-cooked? I wouldn’t.

All the developers must be consulted before agreeing for deadlines. There should be no room for compromise here.

4. Bad perception of development

Big Co. runs its code through all sort of static analysis tools, leak detectors, profilers and what nots, in the hope that tools are going to fish out bugs early. That is a very nice strategy. But what Big Co. fails to understand is that all code advisory tools are just supplementary to sound code. No tool can substitute clear thinking and a taste for programming. Coding experience is a perk but not compulsory. Why can’t Big Co. trust tools entirely? Because the tools themselves are written by Big Co. stereotypes. I bet that if a tool(lets call it MagicWand) comes out in the market which pumps out code for a given requirement, Big Co. would do a mass downsizing, leaving only managers trained in using “MagicWand“. Their only budget allocation would be buying of “MagicWand” licenses. Lesson learnt: there is no silver bullet.

5. Hypothetical targets

Big Co’s managers badly need a crash course in statistics or programming, or in most cases both. They set a very low defect goal, and they estimate code size like 12 yr old kids do. To enrich their much needed estimation know-how, they study books on code estimation(if a book teaches them how to swim, they’d read it as well and dive into the waters!). The importance given to numbers like defect count is blown out of proportion. Statistics are like bikinis, they show everything, except what one would actually want to see.

Adding to the fact is poor measurement techniques. Metrics and timesheets are crappy in Big Co. Developers play their part by filling it with junk, in most cases. Folks bank on these numbers and take cut-offs, the result is evident. rescuetime and ididwork are good a good cure to the metrics problem.

6. Hiring bad people

If Big Co. hires good people, problem #4 would be solved to a great extent. What does “good” people mean? It exactly means people who get things done. If I was hiring for Big Co, I would not go for buzz words or college pedigrees or by asking tweaky questions for which you can find out answers in minutes by googling around. Things which would count are projects worked on(at least one open source), expertise in one programming language, knowing many programming languages and communication. Communication strictly means the ability to document stuff. Code must be written for humans to understand and only incidentally for computers to execute. Big Co.’s managers are tempted to hire more people if the project is running a tight deadline. I would recommend them to read “the mythical man month” by Fred Brooks. No excuses here.

Even if Big Co hires people on a needy basis, their mentoring program sucks. There is no concrete material or approach. Both the mentor and mentee fill a lot of documents and end up doing nothing. No company gets a second chance to impress their employees.

Software development is an iterative process. It is one of the most creative jobs. Developers burn the midnight oil, eat junk food, lose sleep, in some cases, marriage and develop software. They invest more than time in their products. A little encouragement is always great. Big Co must develop a broader perspective on this, failing which we will continue delivering substandard solutions, and getting substandard solutions.

what would be an ideal cluster at home?

badrix — Fri, 18 Jul 2008 13:49:16 +0000

First for the big players in the field.

Amazon’s elastic compute cloud 2 offers servers on demand. Their resizing option is good as well.
http://www.tsunamictechnologies.com/ – provides computing on demand, which closely resembles the model I thought of. This is the pick of the whole lot. They have a lot of information which the “casual reader” might be interested in, like typical applications which one might run, how the service works etc. The parallel python supported by Tsunamic is a honking great idea.
Others who provide computing power for rendering liker render rocket, renderarm(this one I figure is from India, aims to be the biggest in Asia).

I haven’t actually assembled any system yet, because of the power considerations(many people have warned me of this problem, thanks a ton!). I thought I’d rather run all the benchmark programs in any of the above services.

I toyed around with the idea of parallelizing octave, found a good implementation, viz MPI toolbox for octave. This one is essentially all the MPI routines covered up under .oct files. Their website has all the meticulous details required for running a system. They have a ready made cluster OS solution called PelicanHPC which comes with MPITB.

An optimal solution for a mediocre cluster, computing wise and cost wise would be PS3. These folks have found an interesting use for such a behemoth machine. Why would an 8 core processor run with 256 MB of main RAM is a puzzle to me. Secondly, we are devoid of using infiniband($$) for hooking up all the clusters.

Still the fortran/C++ thing is murky. C++ has done a lot of catching up in numeric computing by introducing concepts like template meta programming. Fortran is still preferred partly because it is a legacy language IMHO.

a supercomputer at home

badrix — Mon, 14 Jul 2008 03:51:53 +0000

problem 1. I have too many old computers at home and I don’t know what to do about it.

problem 2. I require a lot of number crunching now and then. time is a premium.

problem 3. I need at least 2 interesting problems to work on. A predictable life is the most boring creation of God ever.

After an unsuccessful attempt to boot linux from scratch rescue system on a P3 machine and a couple of google searches about cluster computers, a familiar tone ringed on my head. Why can’t I solve the problem for a lot of people. People who need lots of computing power momentarily, but can’t afford infrastructure. This was not the first time I get entrepreneurial ideas. Something impulsive strikes time and again, but fades away as soon as it came. I wanted to capture this moment. Nothing dies faster than a new idea in a closed mind. So, I decided to rope in a microcluster out of what I have.

Blue gene from IBM

Currently it sums up to 2 P3s and 1P4. Fair enough. The list might grow depending on the number of people at ebay who are willing to sell computers at throwaway prices.

who would want a sumpercomputer you might ask? I have answers. Using it for parallel computations and large scale simulations is just a niche of the academia and military. I want to compile hell a lot of code in a matter of minutes. I want to render or raytrace a several million polygon mesh without my computer stammering a bit. I want to do FFT and signal processing at a phenomenal rate. The uses just multiply as the number of problems. Otherwise there wouldn’t be as much as 500 supercomputers in the world.

A homwbrew cluster

So, people would be charged on a timely basis, or a gigahertz basis based on feasibility and they would login to their accounts using some RDP(back to this part shortly), and do whatever they want. Some folks might be developers, so they can use their account for developing MPI based applications.

I thought who would buy the idea. Whoever does it needs a proof of concept or at least a working prototype. Not a bunch of promises. Besides, its too much work for one person. I decided to apply for Y combinator. The odds of the proposal being accepted are rare, but I still had something to expect. Their application form was interesting. I have read many of Paul Graham’s essays on startups, but I had a lot of surprises while filling the form. Some questions intrigued me. I didn’t have any cofounder. This was a spot of bother. Another question was regarding open source projects I’ve worked on. I had’nt worked on any. Just used a lot. I decided to roll my sleeves and work on some open source project to materialize this thing.

The next question is, what to work on? Here I had to scratch my personal itch. If I am building a number cruncher, I will be writing number cruncher software for sure. I thought of Octave. It had all the elements I had to work on in the right mix. Numerical computing, compiler technologies, C++, fortran(believe it or not, they use this language a lot in supercomputers, which I learnt only yesterday) and the like.

Where sould I start tidying up here? The answer was easy. Start fixing bugs. For now, my idea stands here. While I fix bugs in Octave, I am targetting on a lot of applications which involve a fair amount of computing power. VLSI synthesis, protein folding, algorithms, prime numbers, rendering for movies to name a few. But as I told, it would be too much work for me, at least for now till I start working on this full time. Some expert advice(talk is cheap, show me the code) and contributions would be great. Even great enough to fill the cofounder slot. Whatever non trivial progress I make will be logged here till its done.