Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror

Comment: ChromeOS + SSH client (Score 1) 99

The Secure Shell extension for Chrome gets you a usable terminal and SSH client. If you need to edit files locally, vim is available too, along with several text editors. There are a couple of git clients floating around if you want to edit stuff with source control.

This isn't such a good option for me. I use LaTeX on my chromebook and want to compile stuff. I want to use sed and awk and wc. I currently use Crouton, though I'm probably going to install Chromixium instead.

Comment: Simulated annealing (Score 1) 140

by dhasenan (#49457579) Attached to: Finding an Optimal Keyboard Layout For Swype

In the T9 section we employed a random walk optimization. For the swipe optimization we use a similar approach but gradually reduce the number of random swaps over time so that the keyboard settles into a local minimum.

A random walk with hops being shortened over time is called "simulated annealing". It's an alternative to genetic algorithms and tends to be easier to use for problems with solutions that can't be chopped up and put together in a coherent format. For instance, keyboard layouts, which require each key to be present exactly once.

Comment: Misleading summary (Score 1) 185

by dhasenan (#49435035) Attached to: The Key To Interviewing At Google

Some interviewers might use "how would you move Mt Fuji" type questions, but, the Wired excerpt explains, these questions and their answers are removed from consideration when determining whether to extend an offer, and the official (and unofficial) policy is not to ask that sort of question.

Nice try, though. The error probably comes from summarizing a summary of an excerpt rather than going to the original source, or at least the full excerpt.

Comment: Re:It's the CSS and JS situation again. (Score 2) 31

by dhasenan (#49381901) Attached to: Developer of 'Banished' Develops His Own Shading Language

For my purposes, I use third-party engines that already can render to OpenGL and Direct3D. It is trivial for me to support both -- unless I have shaders, in which case I have to write my shaders twice. I don't mind doing that for a small number of shaders. I do mind doing that for five dozen shaders.

In the case of SRSL, I think the idea was partly just because he could, and partly because he found himself modifying shaders moderately frequently.

Comment: Re:It's the CSS and JS situation again. (Score 0) 31

by dhasenan (#49380071) Attached to: Developer of 'Banished' Develops His Own Shading Language

You have to write your shaders twice if you want to support Direct3D and OpenGL. That's the biggest problem. A transpiler from GLSL to HLSL or vice versa would solve this, of course.

Direct3D shaders have a bytecode format that is at least somewhat documented (https://msdn.microsoft.com/en-us/library/windows/hardware/ff552891%28v=vs.85%29.aspx), so you could in theory write a compiler from GLSL to D3D shader codes. This might be easier than writing a transpiler in general, though obviously if you're more familiar with GLSL and HLSL you'd find it easier to compile to shader languages than to a bytecode you are unfamiliar with.

Comment: Re:So Red Hat and Ubuntu offer signed binaries (Score 2) 362

by dhasenan (#49307031) Attached to: OEMs Allowed To Lock Secure Boot In Windows 10 Computers

You use a shim bootloader (or GRUB) signed by Canonical, which optionally loads whichever bootloader you normally use. This gives you no security benefits and only serves as a workaround for secure boot.

Granted, this only works for hardware vendors who work with Canonical (or Red Hat or what have you) and relies on them producing a bootloader that works with the operating system you wish to use. GRUB supports the "chainloader" command, but it's possible that hardware vendors might force Canonical to disable this with their signed binaries.

Canonical has spoken about the possibility of distributing a shim bootloader signed with Microsoft's key, too.

All of these are workarounds that make UEFI security worthless. It's better to be able to turn off security or manually import a key than to use a bootloader that will happily load anything and is signed with the same key that restrictive bootloaders are.

Comment: Re:HOWTO (Score 1) 1081

by dhasenan (#49261995) Attached to: How To Execute People In the 21st Century

> Even if you have the right person, it's not actually punishing HIM (or her,) since death is the ultimate fate of all living organisms.

Then you don't mind if I execute you tonight?

The point isn't that the person has to die; it's that they die, presumably, sooner than they otherwise would, with their crime being cited as the motivation for their early death. They are deprived of years of life, just as a person imprisoned is deprived of years of freedom.

Comment: Re:It makes it easy to support "not enough skilled (Score 1) 292

by dhasenan (#49219399) Attached to: Do Tech Companies Ask For Way Too Much From Job Candidates?

It also lets them reject people for arbitrary reasons, citing insufficient qualifications as the overt cause. After all, you tend to lose lawsuits if you tell people that you're rejecting them because they mentioned they're married to someone of their own gender, or because the person is capable of becoming pregnant, or their skin color is an unpopular one.

Comment: Re:What's TSYNC ? (Score 2) 338

by dhasenan (#49210495) Attached to: Google Chrome Requires TSYNC Support Under Linux

On the assumption that a developer is trying to create an application for this sandbox environment, they get a very fast indication that they did something that isn't allowed instead of potentially mysterious errors. On the assumption that you're running malicious code, it means that that code can't continue probing your system.

It does mean that portable code can't probe for what features are enabled and you instead must tell it in advance. Ideally there would be a way to query for which APIs are allowed and which aren't.

Comment: Re:One thing for sure (Score 1) 531

by dhasenan (#49144833) Attached to: Machine Intelligence and Religion

Religions served to concentrate wealth, historically. Furthermore, several of them offered strong reasons to tend the sick, investigate the natural world, and so forth -- probably more so than a typical local lord would have.

Religions also served to unite farflung lands. A researcher in Isfahan could correspond with one in Marrakech. A monk studying flowers in Edinburgh could share his findings with a nun in Osel. Without religions, you need another set of well-funded institutions with a tradition of correspondence and interaction to provide the same benefits. Today we have universities and research organizations like Oxford and Brookhaven National Labs. The concept of a university grew out of the Muslim monastic tradition, though, starting with the University of al-Qarawiyyin, founded in the 800s.

Of course, you can still get the same benefits without religions. But we have no reason to think we'd be even this advanced without the monastic traditions of Islam and Christianity.

There are three kinds of people: men, women, and unix.

Working...