4 ms·

This begs the question: Is it healthy that one of the most important technologies today (web) is also very complicated (=expensive) and thus effectively limiti

This begs the question: Is it healthy that one of the most important technologies today (web) is also very complicated (=expensive) and thus effectively limiting number of players.

roca 9y ago
It's not healthy, but unfortunately maintaining compatibility and staying abreast of modern platform features (so that apps and content don't fully migrate to the single-vendor platforms) just mandates a lot of complexity.
- edejong 9y ago
  The reason why it’s expensive is that no HTML5 proposals are taking into account the cost of development and maintenance. “Anything goes” as long as it competes with the walled-garden platforms. Spending without budgeting nor cost evaluation is similar to going to a casino spending your life savings thinking you’re the lucky one.
  - roca 9y ago
    Deciding just to cede applications and content to single-vendor platforms isn't that appealing for the open Web.
    - edejong 9y ago
      That implies a binary decision. The war is won in many battles. Deciding which ones to go into is the deciding factor. Bringing your men to each is a sure way to lose, and it’s what is currently happening with the modern web.
      - roca 9y ago
        If individual browser vendors decide that some feature is not worth the effort, they can just not implement it and not support those apps fully. Maybe change that decision if the feature gets popular. If the feature never gets popular it can be removed. This actually happens all the time. But vendors getting together and saying "no, we're just not going to compete with native platforms in this space" seldom happens. As a proponent of the open Web, I think that's a good thing.
        edejong 9y ago
        Browser vendors usually do not have this choice. Either they proposed the change themselves, forcing other vendors to follow, or a prisoners-dilema occurs where either all vendors act, or non at all. These are well studied problems in oligopolies and are the prime reason we need a certain amount of governance of (independent) standards bodies.
        roca 9y ago
        Browser vendors always have the choice not to implement, or to delay implementing, standards are are seeing little uptake. I know, I did this work for over a decade and made many such decisions.
Sylos 9y ago
Of course, it's not always healthy, but what are you gonna do about it?
adrianN 9y ago
All of the most important technologies today are very complicated. Anything that has to do with computers, cars, airplanes, trains, almost every industrial process. Being complicated is basically one of the prerequisites for being "high" tech.
- amelius 9y ago
  I think parent meant a different kind of "complicated". More like "overcomplicated". If we want to move forward as a society, we better structure our tech in the simplest possible way. The difference with cars, airplanes, etc is that in a capitalistic world, an overcomplicated browser spec actually helps big players to keep their monopoly. That's what's wrong here.
  - madeofpalk 9y ago
    I'm confused what the move should be then? People want more features and capabilities from the products they use, websites included. What im reading here is a preference for web technologies to be frozen as they were since 2000.
    - amelius 9y ago
      That's a difficult question for which there are probably many answers, so here's my attempt at it. One first insight: we don't need the "semantic web" because Machine Learning can recover semantics. (This is also how Google's webcrawler views the web, so why not apply it to browsers too? People mix up semantics and formatting too often for the "semantic web" to be a useful concept). Another thing to realize is that overcomplicated systems benefit from factorization, i.e. splitting functionalities into different modules with their own responsibilities. This doesn't just apply to code, but also specifications. A better factorization will immediately solve the "monopoly" situation, because now many players can develop many modules, and browser vendors can simply combine module implementations at will. What it will also solve is security issues. A modularized architecture is much easier to keep secure than a system where all functionality has been thrown together on one big heap without structure. You can say that browsers are (if the developers are smart) already structured in a modularized way, but that is not the point, because the internal boundaries of the modules are not openly available in any kind of specification. Personally, I long for a world where formatting a document or showing a video on a computer screen is not considered "high tech". And this might be the way towards it. Note that an "extreme" interpretation of this would be to replace the browser by a simple virtual machine. In that case, every web-developer is their own browser vendor, and they can mix and match existing libraries at will. People will then probably object that the user loses power, and accessibility is lost. But we live in a world of Machine Learning now, so any "structureless" rendering can be easily restructured by an appropriate tool. E.g. an image containing text can be OCR'd and read aloud by a tool. This would also make these tools more robust against formatting hacks. And machine learning can do many more useful things, like removing ads.
ptero 9y ago
It is not healthy, but it is not trivial to fix -- I think building a modern spec compliant browser is expensive (=complicated). One way would be to create a browser architecture from truly independent pieces that would allow one to pick implementations from different sources. That would split complexity and hopefully increase choice. To get this off the ground one would need an architecture and a functional reference implementation, which is a lot of work. If you get that, others will improve it from an OK state to greatness, but it must be end to end OK to start with for people to notice
TheAceOfHearts 9y ago
I'd agree that the browser spec is incredibly complex, but a large part of it is due to not wanting to break compatibility with the web. Many of the problems tackled by the browser are just complicated, and there's no shortcuts. If we look at other technologies, the same thing happens there. How expensive would it be to write a new OS from scratch? A new C++ compiler? A new UI toolkit? These technologies are all incredibly complex. Heck, consider how many projects are just frontends for LLVM. The LLVM IR claims to be well documented [0], but AFAIK it lacks a formal spec as well as multiple independent implementations. [0] http://llvm.org/docs/LangRef.html http://llvm.org/docs/LangRef.html

This begs the question: Is it healthy that one of the most important technologies today (web) is also very complicated (=expensive) and thus effectively limiting number of players.

roca 9y ago

It's not healthy, but unfortunately maintaining compatibility and staying abreast of modern platform features (so that apps and content don't fully migrate to the single-vendor platforms) just mandates a lot of complexity.

edejong 9y ago
The reason why it’s expensive is that no HTML5 proposals are taking into account the cost of development and maintenance. “Anything goes” as long as it competes with the walled-garden platforms. Spending without budgeting nor cost evaluation is similar to going to a casino spending your life savings thinking you’re the lucky one.
- roca 9y ago
  Deciding just to cede applications and content to single-vendor platforms isn't that appealing for the open Web.
  - edejong 9y ago
    That implies a binary decision. The war is won in many battles. Deciding which ones to go into is the deciding factor. Bringing your men to each is a sure way to lose, and it’s what is currently happening with the modern web.
    - roca 9y ago
      If individual browser vendors decide that some feature is not worth the effort, they can just not implement it and not support those apps fully. Maybe change that decision if the feature gets popular. If the feature never gets popular it can be removed. This actually happens all the time. But vendors getting together and saying "no, we're just not going to compete with native platforms in this space" seldom happens. As a proponent of the open Web, I think that's a good thing.
      - edejong 9y ago
        Browser vendors usually do not have this choice. Either they proposed the change themselves, forcing other vendors to follow, or a prisoners-dilema occurs where either all vendors act, or non at all. These are well studied problems in oligopolies and are the prime reason we need a certain amount of governance of (independent) standards bodies.
        roca 9y ago
        Browser vendors always have the choice not to implement, or to delay implementing, standards are are seeing little uptake. I know, I did this work for over a decade and made many such decisions.

Sylos 9y ago

Of course, it's not always healthy, but what are you gonna do about it?

adrianN 9y ago

All of the most important technologies today are very complicated. Anything that has to do with computers, cars, airplanes, trains, almost every industrial process. Being complicated is basically one of the prerequisites for being "high" tech.

amelius 9y ago
I think parent meant a different kind of "complicated". More like "overcomplicated". If we want to move forward as a society, we better structure our tech in the simplest possible way. The difference with cars, airplanes, etc is that in a capitalistic world, an overcomplicated browser spec actually helps big players to keep their monopoly. That's what's wrong here.
- madeofpalk 9y ago
  I'm confused what the move should be then? People want more features and capabilities from the products they use, websites included. What im reading here is a preference for web technologies to be frozen as they were since 2000.
  - amelius 9y ago
    That's a difficult question for which there are probably many answers, so here's my attempt at it. One first insight: we don't need the "semantic web" because Machine Learning can recover semantics. (This is also how Google's webcrawler views the web, so why not apply it to browsers too? People mix up semantics and formatting too often for the "semantic web" to be a useful concept). Another thing to realize is that overcomplicated systems benefit from factorization, i.e. splitting functionalities into different modules with their own responsibilities. This doesn't just apply to code, but also specifications. A better factorization will immediately solve the "monopoly" situation, because now many players can develop many modules, and browser vendors can simply combine module implementations at will. What it will also solve is security issues. A modularized architecture is much easier to keep secure than a system where all functionality has been thrown together on one big heap without structure. You can say that browsers are (if the developers are smart) already structured in a modularized way, but that is not the point, because the internal boundaries of the modules are not openly available in any kind of specification. Personally, I long for a world where formatting a document or showing a video on a computer screen is not considered "high tech". And this might be the way towards it. Note that an "extreme" interpretation of this would be to replace the browser by a simple virtual machine. In that case, every web-developer is their own browser vendor, and they can mix and match existing libraries at will. People will then probably object that the user loses power, and accessibility is lost. But we live in a world of Machine Learning now, so any "structureless" rendering can be easily restructured by an appropriate tool. E.g. an image containing text can be OCR'd and read aloud by a tool. This would also make these tools more robust against formatting hacks. And machine learning can do many more useful things, like removing ads.

ptero 9y ago

It is not healthy, but it is not trivial to fix -- I think building a modern spec compliant browser is expensive (=complicated). One way would be to create a browser architecture from truly independent pieces that would allow one to pick implementations from different sources. That would split complexity and hopefully increase choice. To get this off the ground one would need an architecture and a functional reference implementation, which is a lot of work. If you get that, others will improve it from an OK state to greatness, but it must be end to end OK to start with for people to notice

TheAceOfHearts 9y ago

I'd agree that the browser spec is incredibly complex, but a large part of it is due to not wanting to break compatibility with the web. Many of the problems tackled by the browser are just complicated, and there's no shortcuts. If we look at other technologies, the same thing happens there. How expensive would it be to write a new OS from scratch? A new C++ compiler? A new UI toolkit? These technologies are all incredibly complex. Heck, consider how many projects are just frontends for LLVM. The LLVM IR claims to be well documented [0], but AFAIK it lacks a formal spec as well as multiple independent implementations. [0] http://llvm.org/docs/LangRef.html http://llvm.org/docs/LangRef.html