(this is page 117 of 121)





Latest

It's good to whine. In response to my complaining about how the debate over Python and Ruby, and Rails and Django is getting old, someone sent me this Ian Bicking article comparing Python and Ruby.

Now I know I said I was ready to move on but this piece is just '''so''' good, I had to mention it here. From what little Ruby experience I have, I get the feeling that this is the best overall comparison between the two out there-- and what is really great is how much of the "conventional wisdom" regarding differences he just cuts right through, i.e., Ruby being more "OO" than Python as a common misconception.

In fact the only think I would say is missing (or just not directly addressed) which I would have mentioned is transparency of implementation. For example, in Python all the OO is done with dictionaries which makes it really nice conceptually. I don't have enough experience in Ruby to know whether this is also the case, but when I've played around with re-opening classes I've gotten this weird feeling that there is something much deeper going on inside the Ruby runtime.

Finally, Ian has this fabulous link to a little widget from the computer language shootout data which is just great. It is in desperate need of some Tuft-ean help presenting the data (it takes a while of staring at it to figure out just what it is they are trying to show) but man is it fun to run comparisons like this one and wonder just WTF all those millions of clueless programmers are doing (look across all 3 dimensions), this one to understand just how big the programmer productivity penalty is, or this one to wonder how many times we forget that all the truly genius things in software were done 30 years ago.

Comments Permalink

The Crack that is Productivity

So there have been two raging debates recently that have made me think about productivity and tools.

The first was ignited by one of the Y-Combinator companies, Reddit, reimplementing their site in Python after having done the first version in Lisp. Since they are a "Paul Graham company" and since Paul Graham has spent a long time convincing us thatLisp was the ultimate language (even while speaking at Python conferences), the geek blogosphere and comp.langs.* newgroups were full of debate about what this meant with respect to whether Lisp was all that Lispers make it seem.

The second (but related) debate came from a post on O'Reilly Radar where Tim O'Reilly pointed out that Ruby books have raced past Python books in terms of sales. This observation lit up a debate about the relative merits of the two languages, their web frameworks, and their communities.

These discussions are mildly interesting to follow, but at this point only just mildly. Both Ruby and Python are incredibly productive tools, especially when compared to their C++/Java/C# alternatives. Having slogged through half of a Lisp book, I would imagine that Lisp would also fall into the Ruby/Python camp but for its poor libraries.

So why do I follow the threads at all? Because it was one such thread back in 1995 that led me to Perl, and another one in 1999 that led me to Python. In each of these cases, the deltas in productivity were so great that it was a worthwhile investment to follow what the alpha-geeks were discussing.

It's just that now we're a year into the web frameworks brouhaha and yet it seems like the discussion has stagnated. Yes, Rails is much better than J2EE for productivity with small teams. But Rails and Django are eerily similar in the way they attack their problem (as are the half dozen other web frameworks that have sprung up in the Python world to try to be "the next Rails"). Why is it that almost no one is talking about novel languages/frameworks/approaches that are truly new to the way in which people solve web problems? Where are the approaches that are as different as dynamically-typed, interactive interpreter-enabled languages like Python was to Java six years ago?

I spent some time this last week digging around for what that next novel thing might be. This led me to a fairly mediocre book (which would have been great as a series of blog posts) called Beyond Java which had a chapter at the end about a Smalltalk framework called Seaside which takes a continuations-based approach to handling the processing of web requests. Really interesting stuff.

I'm not sure about whether Seaside is the type of measurable leap forward that I'd like to discover, but I do know that I would like to see some of them energy that is going into arguing about whether Ruby has really eclipsed Python or whether the Reddit rewrite proves that Lisp is just another one of those insanely productive languages (as opposed to The Most Insanely) shifting to discussing the pluses and minuses of languages with continuations support in them.

At the very least, tracking such a debate like this would surely help me in finding my next fix.

Comments Permalink

The Grain of the Web

Though the topic of defining Web 2.0 has been beat to death, Paul Graham throws his hat in the ring with another great essay.

The arguments are thoughtful but the best part comes at the end when he discusses the nature of the medium:

Web 2.0 means using the web as it was meant to be used, and Google does. That's their secret. The web naturally has a certain grain, and Google is aligned with it. That's why their success seems so effortless.

Love that notion. And this essay just goes to show that a fresh perspective can help make two of the most tired, over-written topics in the blogosphere these days (Web 2.0 and Google) interesting again.

Finally, (semi-related) for a little entertaining Thanksgiving reading read the talk that George Dyson recently gave at Google. I won't spoil the end by excerpting but it is quite a fun read.

Comments Permalink

The Culture of Tools

I've spent the week interviewing for a new engineering we've got and have met with some good folks. Among the things I like to talk about is programming language preference. Now I know that all good "computer scientists" are not supposed to care beyond the fact that a language is Turing complete, but the reality is that asking a potential developer about language preference is akin to asking them about their esthetic without sounding pretentious.

One of the things I always try to get to right away when talking about programming languages is the strong-verus-dynamic question. I am a firm believer that late-binding (latent typing) languages like Python, Ruby, and even Perl are one of the best tools for developing products when you don't really know what it is you're going to end up with when you start. A lot of smart people have made this very observation before so I am not going to belabor it here. But I was a bit surprised this week when I heard from someone who I was fully expecting would embrace the dynamism of the Python (by stuff I'd seen on his blog) that the dynamic nature of Python wasn't what really drew him to it.

After giving this a little thought, I realized that he was right in the following sense: the open source languages are just as interesting for the communities they attract (and what ollows from these communities) as for any major features in their design/implementation. I would still contend that latent typing is a Very Good Thing for a lot of application work but I am glad to be reminded that comp.lang.python (and all of the culture that comes with it) can be just as important.

One quick example: among the folks I met this week was a very senior J2EE architect that came highly recommended from a series of very successful venture-backed startups. The guy was clearly very sharp and a true Java expert. During our conversation he let it be known that Aspects were the new Holy Grail for the bleeding edge of Java. Never having really understood the big deal with aspects, I spent most of our hour together quizzing him about the way he had successfully used aspects in his recent work. He gave me some examples and then pointed to some other open source luminaries who are better future architectural directions of aspects.

And yet at the end of the day, all that I came away convinced of was that aspects were a hyped up version of decorators in python. Now we use decorators in our app for a couple of neat hacks but I never would have thought to make a big deal about it because I happened to have read about them one day while following an interesting debate on comp.lang.python.

It's good sometimes to look up and see the forest for the trees.

Comments Permalink

Get Lost

It is a rare pleasure to see something that is so good-- so far above the rest of anything else that might compete with it-- that while you are using it (or watching it), you keep asking youself how it is possible that this thing/service/content could be so good. It happens with Apple and their portables, jetBlue and their flights, and in the case of recent television shows, the first season of Lost.

After watching the first episode for exactly four minutes it became perfectly clear to me that this was a television show cut of a different cloth than just about anything else out there. In it's first season (disclaimer: I have not seen any of the second season) is better written, better shot, better edited, better cast, and better acted than anything else out there. By a long shot.

The obvious question to ask is: why Lost and not the other two dozen shows that have been created over the last two seasons? Or perhaps more strikingly, given Hollywood's "studio model" where the same folks come together in different combinations to work on projects, why is there such a big gap between the quality of Lost and the quality of just about everything else out there?

I'm not sure I know the full answer, and frankly I'm not sure JJ Abrams (the creator of the series) knows for sure either. After all, his previous hit, "Alias," is a far inferior show-- in fact at it's core it's just an hour-long music video with a semi-klad Jennifer Gartner changing costumes and bouncing through raggedy-thin plot lines.

It was so striking though that I went looking for wisdom in the 'Bonus Features" part of the Season 1 DVDs. I don't often get into the bonus tracks mostly because of the way in which they end up making the actors look so much more clueless than the characters they portray, but in the case of Lost I just had to look. And I'm glad I did not so much because of anything that Abrams says but because of the smart things that his producers, set people, and other staff say about him. Here I want to focus on the two observations made that I think can best explain why the show is so good:

  • talent: on the casting side, someone described JJ's style as finding the right people and then re-writing the characters around them. I am not sure how common this is but when you see characters as complex as Kate and Hurley and listen to how their characters were written before the actors were cast (and subsequently re-written), you quickly realize that this was a team that was just not afraid to make wholesale changes to the story to accommodate the way in which the actors could bring different characters to life.

  • creative direction: I've heard this one about great CEOs as well but it's still worth mentioning it here. One of the production people says at one point that what JJ is really good at is having "very clear big concepts" and then letting the team (actors, cinematographers, special-effects people) "texture" the ideas to get to the final product. The "having the big thoughts" bit is part of every bonehead business book written so it's nothing new. What is not often said however is that it's also very important to be crystal clear about the big idea or plan. Without clarity it's very hard to discern when someone's "texturing" is going in the right direction.

This has got me thinking about how to apply these two insights to building software. You can't quite draw a direct parallel because the product delivered in both cases is so different. But it is an interesting thought experiment to think about what would happen if product features could be dictated by the specific talents of the engineers and designers hired. To some degree features are emphasized (or de-emphasized) according to engineers's preferences unofficially, but the whole process of "software engineering" with its product management, functional specs., etc. is in place to correct for this. Maybe the new era of software delivered as service (with much tighter time-to-market schedules and instant feedback from users) will help to get us there.

So, go and get Lost. You won't regret it.

Update: I started writing this post while on an airplane ride to California. While there I met with a bunch of folks at different places including some at Google. It occurred to me after talking to them about their work that Google gets this JJ Abrams insight about shaping the work around the talent. It was pretty incredibly to see a company of 4,000+ people that can operate like this at the scale they are at.

Comments Permalink