> At least chromium is open source.
That source code is 12-15 million lines of code. They implement literally thousands of specs. Chrome adds 40-70 new web APIs [1] with each release which happen roughly once every 40 days and contain god knows how many code changes.
Good luck deducing "precise specs" from that.
> I wish Mozilla or a similar organization adopted the chromium core
That would mean cementing Google's monopoly of the web.
Exactly, it's impossible to describe that in human language which doesn't have precise enough semantics. That's why the spec has to be in code, in the form of a reference implementation and/or test suit.
The goal should be to move chromium away from being controlled too much by Google.
On the contrary, this implements specs that are defined in a human language. For example, the full semantics of rendering HTML were standardised with HTML5, and now all browsers render HTML the same way.
The problem isn't the human language. The problem is the sheer number of these specs. Hundreds, if not thousands.
However, good luck understanding these specs by reading code or unit tests (and yes, there's a multitude of tests [1]). How do you know what a piece of code encodes? Or a test tests? You really propose to reverse engineer behaviour from code?
Example: here's a test: https://github.com/web-platform-tests/wpt/blob/master/html/s...
What exactly does it test? Why? Is this a correct test for this behaviour? Care to pinpoint the exact precise location of this spec in Google Chrome's code? [2] It's only 12 or so million lines of code, should be easy to find.
> The goal should be to move chromium away from being controlled too much by Google.
By doing what exactly?
Is this sarcasm? Writing web apps that work the same across browsers is still the most painful part of web development. The underspecified natural language specs lead to subtly different behaviors in different browsers.
Full specs of a complex systems are very complex.
It's not. How HTML is rendered was literally not defined between browsers until mid-00s. Hence the difference in the boxing model, for example.
Bt yes, besides HTML there are hundreds of other specs, and those may or may not be implemented across browsers.
> The underspecified natural language specs lead to subtly different behaviors in different browsers.
You keep saying this, and I don't think you even understand what you're saying.
> Full specs of a complex systems are very complex.
Yes, they are. What you keep proposing though, is to reverse engineer behaviour from millions of lines of code. I'm still waiting for an answer to my question about datalistoptions test.
The technical blueprints are as important than the natural language notes in the blueprints. In software, code is blueprint.
At least chromium is open source. The best we can hope is maybe that organizations adopting chromium other than Google can wield more influence over it.
As I point out here https://news.ycombinator.com/item?id=27222818 human language is not precise enough to write good specifications. Natural language words are polysemic and contextual. See the meaning of "break" for example: https://www.merriam-webster.com/dictionary/break
Kolmogrov showed that fully specified information distills down to computer programs: https://en.wikipedia.org/wiki/Kolmogorov_complexity, https://en.wikipedia.org/wiki/Minimum_description_length
The ideal language for a specification might be a mix of natural language and code with a test suit, something like Literate Programming (https://en.wikipedia.org/wiki/Literate_programming) or its descendants along with modern pull request based discussions that are tied to version controlled code.
I wish Mozilla or a similar organization adopted the chromium core. We really need a well funded non-profit managed release of the reference browser.