about the existence (or lack of it) of silver
The topic is: are there improvements in programming technology (i.e. silver bullets) that can account for an order of magnitude improvements in programmers productivity.
In other words: if a programmer learned a new technology, could he write an average program an order of magnitude faster?
One technology that is peddled as a silver bullet are higher-level languages, where C/C++ is today considered a low level language, Java a bit more high level and Python even higher level. It seems to be a no-brainer to claim that you can write code much faster in Python than in Java and in Java faster than in C/C++. And often it is true and there even are some studies showing that using small-scale programming exercises.
How well does this theory of Python supreme productivity over C/C++ plays out in the real world?
Today BitTorrent, Inc. announced acquisition or uTorrent BitTorrent client. BitTorrent protocol wasw invented by Bram Cohen (now of BitTorrent, Inc.) and implemented in Python, the uber productive programming language. On Windows, it used UI based on Gtk and later wxWindows. In both cases the executable was large, using a lot of memory and UI was ugly. For example Windows installer for latest 5.0.3 version is 5.9 megabytes.
BitTorrent is also open-source, so at least in theory it should benefit from the open-source magic fairy, contributing improvements and bug fixes.
When BitTorrent, the protocol, became a runaway success, other clients were developed. Those that are popular today are Azeurus (written in Java), BitComet (written in C++) and uTorrent. It’s unclear what language was used to develop uTorrent, but given the (packed) exe is 170 kilobytes, it’s a safe bet that it was written in C or C++, the assembly language of 21st century.
If you use BitTorrent today, you’ll see that almost no-one uses the original BitTorrent client. The people have spoken and switched to Azeurus, BitComet or uTorrent and uTorrent acquisition by BitTorrent, Inc. makes it very clear.
This is the best study of programmer productivity you can come up with: the same basic idea (BitTorrent client) implemented by multiple teams using different technologies.
Free market has decided that the winner of this race is uTorrent, Azeurus and BitComet. The looser is BitTorrent.
Is uTorrent an order of magnitude better than BitTorrent? If you compare functionality, yes in some respects (resource usage) but in general, no (both will do the job of downloading BitTorrent files just fine). But according to only metric that matters i.e. popularity, uTorrent is orders of magnitude better because software tends to have strong network effects i.e. “winner takes all”.
Let’s recap. BitTorrent:
- written in high-level language (Python)
- a huge (years) head-start on other clients
- benefiting from open-source magic fairy
- a resource hog
- lost popularity battle
- written in low-level language (C or C++, most likely)
- tiny and uses little resources
- wins popularity battle
Even if we take into account the fact that uTorrent had it much easier
(it had BitTorrent Python code to study), BitTorrent had a huge
head-start and should still improve at a much faster rate.
But the truth is that a good, dedicated programmer using C/C++ sword will write better apps than another programmer using high-level language like Python.
Another truth is that despite all those “performance doesn’t matter” claims, Python and Ruby and Java are not viable languages for shipping high-quality Windows apps (C# is getting there). Few counter-examples like Azeurus or Eclipse do not a revolution make. Also, the fact that those languages are successfully used in other domains like web apps doesn’t invalidate the point about client applications. Show me a spreadsheet close to Excel in speed and features written in a high-level language.
The only thing that (occasionally) can deliver order of magnitude productivity improvements are humans.
Human being is the only silver bullet so far.
|« memset() considered harmful||•||Sumatra PDF 0.3 released »|