3 ms·
I extract three major points to his argument: 1. The 'web platform' is difficult to write for because there are many different browsers and operating systems.
by jgrahamc 17y ago
I extract three major points to his argument:
1. The 'web platform' is difficult to write for because there are many different browsers and operating systems.
2. The programming language environment for the web is immature and has sharp tools.
3. The people who write web apps are poor quality.
#1 is true enough. It is painful. That why we write tests, use automated systems and are driving for standards. Of course, he could help by persuading Microsoft to play nice with others on web standards.
Also, he's forgotten the days of programming in assembly language where you had to pick the processor and BIOS you were targetting. That was really ugly, but that's where companies like Microsoft got started.
#2 is also true enough. It is an immature environment. But that will change.
#3 is totally true. There are poor engineers working on the web, but that doesn't mean that there will always be poor engineers on it. It's a virtuous circle. The better things get the more the good people will move to it.
It strikes me that he's in real danger of missing an 'innovator's dilemma' situation. Yes, programming for the web is 'crappy' compared to picking a single operating system with very mature advanced tools. That could be the equivalent of the non-hydraulic backhoes and massive hard disks detailed in the Innovator's Dilemma. Perhaps he just hasn't seen the change that's coming, coming.
- bad_user 17y agoPoint #1 is meaningless. You want to have compatibility with major browsers for more market share, otherwise nothing stops you from supporting only IExplorer or Firefox. It's the same situation with desktop applications, although the situation is a little better because of Windows. But, if you don't want your client to download dozens of MB for the latest .NET, then you use .NET 2.0, if want more market share then you start working on a Mac port. If you want to target mobiles, then you start working on IPhone and Symbian ports, while keeping an eye on Android. If it's a game you're developing, then you start working on a port for PS3 / XBox / Wii. The whole allure of web applications is that you can reach a larger audience faster, with no middle-man between you and your customers (no installation required, no manual upgrading to the latest version, no piracy). Your work can be a lot easier when you're using cross-platform tools, but then again, it's the same for browser incompatibilities. Javascript hurts? Then use JQuery or something more heavy like GWT, or maybe OpenLaszlo. CSS hurts? Then use Javascript to simulate missing properties (and the situation is a whole lot better than providing an OS-consistent GUI for both Windows and Mac). I think points #2 and #3 are in conflict, because #3 happens because the barrier to entry is lower. How can that be if the tools we use for the web are immature? His main point is that the web relies on dynamic languages which don't have IDEs like Visual Studio with intellisense / automatic refactoring or compilers that highlight syntax errors at compile-time (well, he actually mentioned HTML as the language, but I'll close my eyes to that). While I can see from where his point comes from, it is also missing the forest from the trees. From my p.o.v. this is a virtue more than a drawback because it makes you think more about your API (your average .NET / Java API design really makes me cry), and it makes you think more about what's hard about the process of development (unit-testing, functional testing, standards, and all that). He also implies the lack of "object-oriented languages" and that of debuggers ... which is a complete fallacy. Modern OOP concepts are not available in C# (mixins, traits, multi-methods), and the debugging in Python (just an example) is a lot more comfortable than what's available in VS.NET. Does VS.NET give you a way to inspect / redefine active objects / methods / entire classes on a breakpoint and then resume execution? And how much do you have to pay to get functionality available for free in Valgrind/KCacheGrind? Also, in platforms with continuations support the debugging takes a whole new perspective. Overall, the article is just a poor and content-free rant.