Slashdot stories can be listened to in audio form via an RSS feed, as read by our own robotic overlord.


Forgot your password?

Comment: Re: DirectX is obsolete (Score 1) 117

by TheRaven64 (#48902995) Attached to: DirectX 12 Lies Dormant Within Microsoft's Recent Windows 10 Update
Actually, a lot of these games just use WINE's implementation of DirectX. This either translates the calls to OpenGL or implements a DirectX state tracker directly if you have Gallium drivers configured correctly. The same is true of a lot of Mac games. Good luck getting WINE to run on a console though...

Comment: Re:DirectX is obsolete (Score 1) 117

by TheRaven64 (#48902991) Attached to: DirectX 12 Lies Dormant Within Microsoft's Recent Windows 10 Update

Your typical GPU driver is about 10MB of object code. It contains a complex optimising compiler and controls a device that has complete DMA access to your computer. It is written with speed as the only significant goal. Making a GPU driver 1% faster contributes enough to sales to pay the salaries of several driver developers. Making the GPU driver more secure generates zero additional sales.

The shader code that's fed into this stack from WebGL is sanitised and is completely safe to run, assuming that your driver stack is 100% bug free. Still feel safe?

Comment: Re:DirectX is obsolete (Score 1) 117

by TheRaven64 (#48902983) Attached to: DirectX 12 Lies Dormant Within Microsoft's Recent Windows 10 Update

If you write a game that uses Direct3D, you can easily target Windows, XBox, and Windows Phone. If you write a game that uses OpenGL, then you can easily target all of the major desktop, mobile, and console platforms. If your game runs on a generation-old console, then it will run on current-generation mobiles as well. This gives you three markets: First release for high-end PCs, second for consoles, third for mobiles. You can get a solid revenue stream out of each one. You don't lose the Windows marked by choosing OpenGL, but you do lose every other market by using Direct3D.

That said, the APIs are so similar these days that you'll typically use some middleware to provide the abstraction. All of the important code is written in the shaders and these are much easier to port between GLSL and HLSL than they are to port between different GPUs and maintain performance.

Comment: Re:Modula-3 FTW! (Score 2) 400

by TheRaven64 (#48902963) Attached to: Ask Slashdot: Is Pascal Underrated?

Implementing this in the standard library means that the language needs to support pass-by-reference (which Pascal and C++ do, but C doesn't). This single feature does a lot to reduce readability. In C, I know that inc(x) is a function that does not modify the value of x, without reading any additional code[1]. In Pascal or C++, I need to look at the definition of inc() to know if x will be the same before and after the call.

An important idea at the core of readability for a language is the amount of code that I have to read to understand a single line. In any language that has pass-by-reference, this amount is larger than a language that doesn't. To achieve the same thing in C, I'd have to write inc(&x), and then everyone reading that code would know that x may be modified. (Note: the almost-equivalence of array and pointer types in C is a good counterexample where Pascal wins massively in readability).

[1] It could be a macro, but most coding conventions require macros that can't be used as if they were functions to be all-caps.

Comment: Re:Modula-3 FTW! (Score 1) 400

by TheRaven64 (#48902951) Attached to: Ask Slashdot: Is Pascal Underrated?
This would be true, except for two things:
  • Lines with more than 66 non-whitespace characters decrease readability.
  • Statements with more than 66 non-whitespace characters are common in most programming languages.

This means that you end up either with lots of continued statements or lots of overly-long lines in Python. If you have the former, then it's hard to see the indentation. If you have the latter, then you can see the indentation but the overall readability suffers. This can be fixed by using tabs for semantic indentation and spaces for alignment and an editor that supports highlighting tabs, but the Python style guides tell you not to do this.

Comment: Re:Modula-3 FTW! (Score 1) 400

by TheRaven64 (#48902941) Attached to: Ask Slashdot: Is Pascal Underrated?

There are some simple things where C is far more readable to a moderately experienced programmer. Consider the beginning and ending of blocks. In pascal, these are signified by begin and end. When you look at a chunk of Pascal code, they can be hard to pick out because they're just words in a sea of words. In C, you use the { and } symbols. These are symmetrical and the human brain has spent a lot of time evolving to be trivially able to spot symmetry because symmetry normally means 'predator about to try to eat me'. You can very quickly spot a column that has a { at the top and a } somewhere later (much more easily if they're aligned together and there's nothing else on the line). There were some studies done in the '80s that confirmed this, though sadly a lot of C coding conventions specify brace placement in a way that reduces readability.

The main strength of Pascal is that it forces you to think more than C. If you don't write what you mean in Pascal, it usually fails to compile. C will happily do... something. This level of redundant verbosity makes Pascal both quite a frustrating language for experienced developers and a great language for teaching. I find that people who learned Pascal tend to write better C code than those that didn't, but neither group has a strong desire to write Pascal.

Comment: Re:Cardholder services (Score 1) 230

by Jaysyn (#48886029) Attached to: Dish Network Violated Do-Not-Call 57 Million Times

Snicker. Another RW moron that doesn't know that Saint Reagan actually began the "Obamaphone" program.

"The Lifeline program originated in 1984, during the administration of Ronald Reagan; it was expanded in 1996, during the administration of Bill Clinton; and its first cellular provider service (SafeLink Wireless) was launched by TracFone in 2008, during the administration of George W. Bush. All of these milestones were passed prior to the advent of the Obama administration."

Comment: Re:Choose a CMS you like (Score 1) 298

by TheRaven64 (#48882861) Attached to: Ask Slashdot: Has the Time Passed For Coding Website from Scratch?

- No security issues (other than bad Javascript or the web server itself): there's nothing to hack, and if someone were to hack the web server itself, restoring the site is as easy as re-uploading the files (all of which can be maintained in version control like git).

For something like Jekyll, this also applies to the input. I use it for a couple of sites and, in both cases, the sources are Markdown files (easy to edit with your favourite text editor) stored in a git repo. When I'm working on updates, I run 'jekyll serve' locally and get a copy of the site on the loopback. When I want to push them, I can do jekyll build and then rsync the results to the web server (or do something more clever if I'm less lazy and want atomic updates). The entire change history of the site is stored in revision control and the revision control system contains everything necessary to recreate the site at any point in its history.

I've yet to see a CMS that allows trivial rollback to earlier versions of the site or which makes it easy to store the content in such a way that a compromised web server can't damage it.

Comment: Re:HTML = programming (Score 1) 298

by TheRaven64 (#48882795) Attached to: Ask Slashdot: Has the Time Passed For Coding Website from Scratch?

I'm not sure I agree with your first premise. There are fairly trivial combinators that you can write in lambda calculus that are conditional flow control (i.e. reduce to either the left or right lambda expression based on a value). The implementation of ifTrue in Smalltalk (loosely) follows this model.

More mundanely, the statement is obviously false because a language constructed with the basic arithmetic operators and unconditional branches is also turing complete.

Only if the unconditional branch is a computed branch. Otherwise how would you implement a program that either terminates or does not terminate based on user input? The example that comes to mind is the x86 MOV instruction which, with a single unconditional backwards branch is Turing complete, but this relies on several other aspects of the surroundings that allow you to implement a conditional branch (or, at least, a select, which is morally equivalent).

The simplest Turing-complete instruction set is a subtract-and-branch-if-not-negative instruction, but this is a conditional branch.

I agree that conditional flow control is slightly too broad a requirement, as it depends on an imperative model. Conditional execution depending on input data might be a better way of phrasing it.

Top Ten Things Overheard At The ANSI C Draft Committee Meetings: (10) Sorry, but that's too useful.