About Vinod Kurup

HTML5 Up and Running

2010-08-25T14:53:00Z

I preordered Mark Pilgrim's HTML5, Up and Running from Amazon back in March. It's available for free as Dive into HTML5. I've been a fan of Mark's writing for a while, so buying the book was a way to thank him "with a buck". Interestingly, the online book and paper book are not identical. For starters, the titles are different. In addition, the beautiful fonts and cutesy images of the online version are gone. The paper book is your run-of-the-mill O'Reilly book. I'm not sure if it'll make much of a difference in my enjoyment, but we'll see.

As I said, I preordered the book in March, so I had no idea when it would be finished and delivered. It showed up on my doorstep this morning, on my birthday! We're on our way to NYC tomorrow for our first trip back there since we moved to North Carolina in 2007, so I know have some light pleasure reading. Perfect Timing!

I've only read the first chapter so far, but I think I'm going to love it. I've done a lot of web programming in my life, but only maintain my own sites now. I'm reading this book like a programmer might read a book on astronomy... purely for the love of it. The first chapter goes into the details of how the IMG tag was invented. Marc Andreessen mentioned it on a mailing list back in 1993, and after some back and forth, shipped it in Mosaic, an early browser. Because it shipped and because it was popular, it got included in the HTML standards.

I still remember when Mosaic came out. I got my first email address in 1992 (vkurup@macc.wisc.edu) and somehow was on a mailing list announcing the impending release of Mosaic. I remember the text of the email basically saying that "This will change the world." Once I saw it, I thought it was cool, but way over-hyped. Who is going to want to wait for pictures to download over a 14.4K modem? Ridiculous. This is why you should not take my advice about new trends.

Anyway, the point of the IMG example is to show that HTML standards have never been "pure", but have always been influenced by the implementers and the users of the web. The way it should be. HTML5 is an acknowledgment of that fact.

Primary Care Doctors

2010-08-06T11:59:00Z

I used to be a primary care physician, but switched to being a hospitalist in 2007. I loved parts about being a PCP, namely the relationship that I developed with the patient over many years. I miss that. I think PCPs offer huge value in providing low cost, quality care, especially if they can stay engaged throughout the hospitalization process. Atul Gawande recently wrote a powerful article on the Hospice system. It's worth reading to being to understand the complex issues involved in end-of-life care, not to mention for the emotional storytelling that is his strength. But, to me, the unsung hero in that story was Chuck Morris, the primary care physician, who came in at the proper moment and provided perspective to the patient that he knew so well, allowing them to make the sane, yet horribly difficult decision to pursue a palliative path. Reading that made me miss primary care medicine a little. I perform much more actual "medical" work as a hospitalist than I ever did as a PCP. But, when I think of the concept of a doctor, in my head, I'm thinking of a primary care doc.

Maps are functions, too

2010-08-05T15:16:00Z

Braindump time. Here's what I've read about clojure in the past few days. I don't understand what this means yet, but maybe someday I will:

Clojure provides easy access to the Java frameworks, with optional type hints and type inference, to ensure that calls to Java can avoid reflection.

There are multiple languages that target the JVM (besides Java, of course). It seems that there are 2 classes of languages. Ported languages such as JRuby and Jython aim to be identical to their base languages (Ruby & Python). Native languages such as Clojure and Groovy are meant to target the JVM natively without satisfying any desire to be identical to another language. Native languages have the advantage that they only have 1 set of libraries to use, so you don't get stuck between two languages.

I installed clojure 1.0 from the ubuntu repositories, but promptly uninstalled it once I found leiningen. Leiningen is a build tool for clojure. You can easily create a new project and specify its dependencies and leiningen will download those dependencies, including a full version of clojure for each project. You can have 1 project using leiningen 1.0 and another using version 1.2 sitting right next to it. That seems smart!

I installed clojure-mode on emacs.

Right now, I'm working my way through the labrepl project. It's an interactive tutorial and REPL (read-eval-print-loop) where you can run the examples in the tutotial.

It's always important to know how to find help:

(doc blah) gives documentation about blah.
(find-doc "blah") searches for documentation.
(javadoc blah) opens Java docs in your browser
(show java.util.Date) shows method signatures

Here's something cool: Maps are associative arrays. Here's a simple one:

(def me {:first-name "Vinod", :last-name "Kurup"})

To get the data, you use the get function, like you would in other languages:

(get me :first-name) -> "Vinod"

But, maps and keywords are functions as well, so you can do:

(me :first-name) -> "Vinod"
(:last-name me) -> "Kurup"

Ubuntu Bluetooth Workaround

2010-08-02T21:40:00Z

I use bluetooth to transfer podcasts from my laptop to my Palm Centro. A Palm Centro is like an iPhone or a Droid, except less smart-like and more phone-like. This worked beautifully but broke when I upgraded to Ubuntu 10.04 (Lucid Lynx). There are 2 bugs involved. First, the bluetooth daemon (/usr/sbin/bluetoothd) never gets started. Second, there's no easy way to start the file transfer daemon automatically (/usr/bin/dund). This used to work automatically, but now it doesn't. I've tried to fiddle with udev rules, but have been unsuccessful. I've reported the 2 bugs, but haven't had any response yet. (bluetoothd bug) (dund bug)

My workaround is a shell script which starts those those daemons if they are not running:

vk-bluetooth-start.sh

#!/bin/bash

BLUETOOTH='bluetoothd'
DUN='dund'

# first start bluetoothd
if ps ax | grep -v grep | grep $BLUETOOTH > /dev/null
then
    echo "$BLUETOOTH is running."
else
    echo -n "Starting $BLUETOOTH..."
    /usr/bin/sudo /usr/sbin/bluetoothd -u &> /dev/null
    echo "done."
fi

# then start dund
if ps ax | grep -v grep | grep $DUN > /dev/null
then
    echo "$DUN is running."
else
    echo -n "Starting $DUN..."
    /usr/bin/sudo /usr/bin/dund --listen --persist --msdun call centro
    echo "done."
fi

And here's my script which transfers the files to my phone. It uses pilot-link, which I had to compile from source due to a bug in the way that the Ubuntu version addresses large SD cards. This should be fixed in the next version of Ubuntu.

vk-transfer-podcasts.sh

#!/bin/bash

output=$(pilot-xfer -p net:any -D /Kingston/AUDIO -i *.mp3 2>&1)
errors=$(echo $output | grep -c "ERROR:")

if [ $errors -gt 0 ] ; then
    echo "failure (count: $errors)"
    echo $output
else
    echo -n "success, removing files from laptop..."
    /bin/rm -I $HOME/*.mp3
    echo "done."
fi

If you're wondering why I use a slow protocol like bluetooth to transfer large files to an SD card, rather than just ejecting the SD card and inserting it into the laptop, well... you haven't tried to access the SD card on a Palm Centro. #paininthe*&^@!

Hospitalist Notes 1

2010-07-29T13:50:00Z

My last night shift was a little less busy than most, so I forced myself to pick something about each patient I admitted and look it up. I tend to do this anyway, but I never keep notes or write about it. Here's what I wrote about that night:

Does Buspar (buspirone) need to be tapered?

Answer: No. The full report on Micromedex was a little conflicting, because it stated that there may be some withdrawal symptoms from Buspar, but this quote in the same profile suggests otherwise:

Data suggest that buspirone does not cause physical dependence and as such, no withdrawal effects have been noted (Rickels et al, 1988a; Tyrer et al, 1985; Cole et al, 1982).

Staging COPD

I always forget the cutoffs for the stages of COPD. There are 2 different guidleines - GOLD and ATS/ARS, but fortunately they are almost identical:

Patients need to have an FEV1/FVC ratio < 0.70. Staging is then based on the FEV1.

Stage I (Mild): FEV1 > 80% predicted
Stage II (Moderate): FEV1 50-80% predicted
Stage III (Severe): FEV1 30-50% predicted
Stage IV: (Very Severe): FEV1 < 30% predicted, or < 50% with signs of chronic respiratory failure

Reference: Annals

How to determine calorie level of ADA diet?

We always put diabetic patients on an ADA diet and I choose the calorie level by gestalt. I figured there must be some formula to calculate the calorie level based on the patient's weight. Looking it up, I found a different answer altogether. The ADA no longer recommends an ADA diet in the hospital. Instead, they recommend a consistent-carbohydrate diet. In long-term care facilities (i.e. nursing homes), they go a step further and recommend a 'Regular' diet with consistency in the amount and timing of carbohydrate. They specifically state that there is no evidence to support the classic 'No concentrated sweets' diet. These guidelines are from 2007, but they obviously have not been widely disseminated yet.

It is recommended that the term “ADA diet” no longer be used, since the ADA no longer endorses a single nutrition prescription or percentages of macronutrients.

Reference: Diabetes Care

My Anti-Writing Cycle

2010-07-27T21:30:00Z

I blogged daily for a 6 day stretch a couple weeks ago, which is the longest continuous stretch I've ever done. It was stimulating and stressful at the same time. I haven't heard from the Pulitzer committee about any of the posts yet, but it's not the quality that matters. Writing is more about the process than the product, and I know from experience that writing helps me think. And while most of the posts were written for my own benefit, I really enjoyed the response that I got from family and friends about Playdoh. I think I've read that post 100 times and I still laugh out loud at the end of it. That's the post that started the stretch off. As soon as I had put Kavi to bed, I knew that I wanted to share that story. I wrote it down quickly, so I wouldn't forget it. That writing process stimulated me to want to write more, so I made it an internal goal to write a post daily.

Then I had two night shifts and the streak was broken. I've had a few days off, but I still haven't written anything. I'm still learning clojure, but I fear writing about something that I know nothing about. Even though I've heard others say that this is the best time to write about a topic. I've also constructed a vicious cycle in my head. I want to learn clojure. I want to write about what I learn. I start doing so, but then I skip a day or so. Now, when I have free time, I think about learning clojure, but then I also think about the fact that I should write about it. And since I'm already behind, I don't want to get further behind, so maybe I shouldn't read that interesting tutorial until I've caught up on writing about what I've learned. Next thing I know, I'm monitoring Twitter streams second-by-second for imgur links.

I'm hoping that writing about my mental blocks will help me overcome them. I plan to write more frequently, mostly for my own benefit, but I'm not going to let it get in the way of learning cool, interesting stuff. If my writing is disjointed, so be it. I can fill in the gaps later if anyone cares.

My video editing process

2010-07-20T20:51:00Z

I use kino to edit the videos that we take of the kids. I haven't found a Linux program which was as nice as iMovie was on the Mac, but kino is pretty good. It's easy to splice and reconnect video clips together. It has nice special effects, though my needs are pretty simple. It's able to export in plenty of formats, possibly too many, for my simple mind. But, I've figured it out and have a stable process for getting video from camera to youtube.

I have 3 camcorders - a Flip, my Canon digital camera, and a Panasonic DVD camcorder. I can just copy the AVI files directly off the Flip and the Canon's SD card. The Panasonic records all of the video clips into a single file with a .VRO extension. I found a great utility called dvd-vr, which reads that file and splits it into individual AVI files. Once I have a directory full of AVI files from the various camcorders, I use emacs to rename them consistently. I open up 2 buffers: an editable dired buffer (wdired-change-to-wdired-mode) of the movies directory, and a shell buffer in that same directory. I run a macro which takes the first file in the directory, plays it via an mplayer command in the shell buffer and then renames the file using the date/timestamp of the file. The macro leaves the cursor in a spot where I can enter a brief description of the video and then run the macro again to do the next video. It works pretty well. I then batch convert them to DV format, which is what Kino accepts. Finally, I launch Kino and edit the files, add titles, export them to AVI again, and then upload them to youtube.

I did this today and ran into a problem with kino. The video played at breakneck speed and there was no audio. Google helped me figure out that this was a problem between PulseAudio and kino. Kino is no longer actively developed. The lead developer considers it stable and finished and he's moved on to other endeavors, which is perfectly understandable. He's however, taken the time to describe a workaround. He recommends running padsp kino which routes all audio requests from kino to pulseaudio. This worked, but the audio had a lot of feedback. I found an even better workaround on the ubuntu forums. Use pasuspender kino, which suspends all PulseAudio activity, gives kino direct access to audio devices, and then resumes PulseAudio once kino has quit. This worked with perfect audio quality, though of course, audio from any other program gets muted completely.

I am still on the lookout for a better video editing solution on Linux. Eventually, I'd like to make DVDs of the hundreds of little clips that I have. They don't need to be fancy, but I would like to have some basic menus and they need to work on standard DVD players. I've started doing a little research on this, so will post once I've tried some things out.

Stock Market Timing

2010-07-19T16:08:00Z

If you asked me a few years ago, I would have said you absolutely can't time the stock market. The 2008 crash hurt enough to make me review my convictions a little. I still believe that accurate market timing is somewhere between difficult and impossible. Trying to analyze stock charts to figure out what will happen in the short term is mostly a crap shoot. The way to win in the stock market is to buy value stocks, especially when everyone else is selling.

I do, however, now believe that there is a place for market timing in deciding when to be in the stock market, in the first place. I think there are some trends that have repeatedly predicted poor stock performance. There are some super smart people that have analyzed these trends over at the Motley Fool message boards. They've come up with many indicators that supposedly tell you when to get in and out of the stock market, but my favorite is the 99 day rule ¹.

The basic premise of the 99 day rule is that when the S&P 500 Index stops making new highs, investors get pessimistic and stocks fall. When it starts making new highs again, optimism takes over and bull markets start. The 99 day rule has 2 parts and coincidentally uses 99 days as the cut-off for both parts. It looks at whether a new high has been made recently. It defines "high" as a 99-day high. It defines "recently" as 99 days. So, it looks for a new 99-day high within the last 99 days. The number is arbitrary. Hop over the the Motley Fool boards to see how the dates have been tuned and pick different ones, if you like.

I like this rule, because it's simple, easy to calculate and doesn't have many "signals", so you're not constantly buying and selling. Of course, it's not perfect, but it would have gotten you out of most of the major bear markets.

Here is how the rule has done over time:

Period        S&P 500   Switching   Improvement 
---------     -------   ---------   -----------
1930-1935     -12.2%       4.3%       16.4% 
1935-1940      10.3%      16.2%        6.0% 
1940-1945       7.4%       5.6%       -1.8% 
1945-1950      10.2%       7.4%       -2.8% 
1950-1955      23.6%      22.5%       -1.1% 
1955-1960      15.2%      13.8%       -1.4% 
1960-1965      10.7%      11.7%        1.0% 
1965-1970       5.0%       7.1%        2.2% 
1970-1975      -2.4%       6.8%        9.2% 
1975-1980      14.8%      12.0%       -2.8% 
1980-1985      14.8%      20.1%        5.2% 
1985-1990      20.4%      18.8%       -1.7% 
1990-1995       8.7%       6.7%       -2.0% 
1995-2000      28.6%      27.9%       -0.7% 
2000-2005      -2.3%       4.3%        6.6% 
2005-early2008  2.0%       4.5%        2.5%

You may have gotten to this point and are now wondering why it's worth following this rule. In at least 3 of the past 5 periods, this signal would have lost money versus staying in the market. That is true. The key is to understand that there is no free lunch. Any rule that tries to get you out of major bear markets will be imperfect and will often leave you with less money than if you had chosen to act on a different date (or not act at all). The overall goal is not to increase your returns, but to decrease your risk without sacrificing your returns. This signal does that. Overall, it does actually increase your returns by about 2% annually since 1930.

The other nice feature is that it doesn't have a "slam-dunk" feel to it. My experience is that things that work in the stock market have to have a little of a "gut wrenching" feel to them. If it sounds too good to be true, it is. This rule feels correct, improves returns over the long run, is backtested well, but does so at the risk of some lower performing years. That feels a little gut-wrenching to me.

Here is the author's (mungofitch) description:

By using the rule that I described above, you would have had a total return of 11.80%, with a total risk of 6.71%. So, you would have had 2.3% per year higher returns on average, while experiencing only about 56.9% of the risk. Higher returns, lower risk. That sounds good, right?

But, is this one of those iffy timing systems that really only works because it managed to avoid the crash of 1987 of something? No, actually it has you in the market in the 1987 crash: it's not perfect. But does it really add value on average through the years? Not always, but pretty darned well. Sometimes you do a little worse, but never a lot worse. And when you do better, you do a LOT better. So, overall, it's worth it.

In order to keep track of the signal, I wrote a little python program to calculate whether we are in buy mode or sell mode. It uses the excellent ystockquote module written by Corey Goldberg.

A much more detailed description of the rule and its background.↩

First Steps with Clojure

2010-07-18T16:29:00Z

As mentioned yesterday, I'm teaching myself clojure. I started by trying to install it.

vinod@ike$ aptitude install clojure

Ubuntu has packaged version 1.0.0 of clojure. I always default to the OS-installed version of programs, just for ease of maintenance. The current stable version of clojure is 1.1 and it appears that 1.2 is in beta. I have no idea what has been changed in 1.1 or 1.2, but I'm going to try to get along using 1.0.0 and update only if I need to for a specific reason.

I then read through the tutorial for non-lisp programmers, which succinctly describes the basics of the language.¹ Everything in lisp is either an atom or a list. Atoms include numbers, booleans, strings, symbols, keywords and the nil value. Lists (delimited by parentheses ()) are the basic data structure, but include representations such as vectors (delimited by square brackets []) and maps (delimited by curly brackets {}). Clojure programs are themselves simply lists, which leads to the power of lisp-like languages because code and data are interchangeable.

You use def to create variables and defn to create functions. The language includes loops and conditional statements, just like any other useful language. Only false and nil are false in clojure; zero (0), and the empty string ("") are true, unlike a lot of languages.

The interesting part is the integration with Java. To instantiate a java object, simply call new:

user=> (new java.util.Date)
#<Date Sun Jul 18 16:20:54 EDT 2010>

To call object methods or get instance/class variables, use the dot "." method.

user=> (. (new java.util.Date) (toString))
"Sun Jul 18 16:24:09 EDT 2010"

user=> (. Integer MAX_VALUE)
2147483647

The article contains this interesting quote:

Sequences are in a sense, the core of idiomatic Clojure programming. Understand sequences and the forms that work with them, and you will have cleared one of the biggest hurdles in writing significant Clojure programs.

So, add that to my to-do list. Where to go next? A StackOverflow thread has pointed to me to a few options. I'll probably read one of these next:

I later read that this tutorial is out of date, so I may have to "unlearn" things later. It still comes up as the first hit on Google for "clojure tutorial"↩

Getting some clojure

2010-07-17T19:55:00Z

Steve Yegge is one of my favorite bloggers. I read "You Should Write Blogs" a long time ago and it is one of the main reasons that I still try to keep this blog going at all. He recently took a blogging hiatus and in his return post, he left a little hint that he was interested in a programming language called clojure.

My programming skills have been rusting away (Darn those patients, always getting sick and stuff!), so I've been looking for a little inspiration. I'm hoping that clojure is it. I've always wanted to learn lisp. I know a little elisp and scheme, but I've never gotten comfortable with them. Clojure compiles code that runs on a JVM (java virtual machine). This gives you a lot of the "power" of java, namely the extensive libraries and the widespread platforms on which it will run. But instead of having to write java, you get to write in a dynamic, lisp-like language. At least that's how I understand it.

Specifically, I hope that I'll eventually be able to create applications that run on android phones, which is a java environment. Smartphones are clearly going to be a huge part of the future, if not the way of the future. I don't see myself ever buying an iPhone and Mala already has a Droid Eris that I'd be able to play with if I needed. (Didn't tell you that, did I Mala?) So... Android it is.

Starting today, I'm going to start reading about clojure and taking notes (and maybe even posting them!). I've never been very good about taking notes, ever in my life. But, I clearly need to learn.