Comment Re:Hmmm (Score 1) 205
We had that "pull-down mirror" in our 2008 bottom-of-the-line Sienna. I called it the "bratfinder", at the time.
We had that "pull-down mirror" in our 2008 bottom-of-the-line Sienna. I called it the "bratfinder", at the time.
There's the snarky answer, and what I suspect is the real answer.
First, systemd and everything associated with is just so kewl and shiny that's it's a privilege to even use any of it, which makes it all the more amazing that they're actually welcoming us to do so, instead of making us fight for a place in line.
Second, X11 goes way back before anyone was really concerned with security. I suspect from a core competence point of view, the X11 coders are far more comfortable and far more engaged with the graphical display code than the input side. I get the impression that a lot of effort was spent in properly cleaning and separating the root-requiring functionality. I know I've read of KMS and DRI work for years now. It's been a long road, and I believe it may have only been in the past year that the display side has gotten to the point where they could think about going rootless.
I also suspect that the input device part is not their core competence - they'd like events coming in from "elsewhere" and get back to their graphics work. So along comes systemd, saying, "We'll handle the gnarly details of console access and security for you," and X said OK, if only in the spirit of modularity and going back to their graphics work. (Graphics work includes processing the inputs, not just drawing outputs - I think they'd just like the inputs to be clean and handed to them.)
I've also found that sadly enough, there are plenty of people around a big company who are really good at appearing essential, while really doing nothing themselves and in fact are very good at creating work for others. Unfortunately they also tend to get retained through job cuts, because they appear so essential.
Though I work in a big company we generally manged to have a small, well-focused team. That makes it a good place to work, as long as you can keep your head down, have fun, and not see the chaos and decay around you.
OK, I'll feed the troll. Either X or an X wrapper is suid root. Find the right hole in X, and you've got root. I presume that X or an X wrapper tries to do the best it can, drops capabilities, etc. But it would still be better to not be root at all.
Are you able to explain more?
My impression is that there were 2 issues with non-root X - mode setup and input device management. KMS and DRI2/DRI3 take care of the former, and I'm under the impression that systemd-logind takes care of the latter. But ultimately these are all just kernel interfaces - if systemd-logind has a root-helper and makes a series of kernel calls to manage the input devices, then that same job could be done by some other piece of software.
Again, do you understand the base mechanism at work here?
As we better understand the universe, we find gaps between reality and our understanding. We then try to extend our understanding to better match reality, and that means filling in those gaps. Sometimes it takes many tries to fill in a gap, or at least make it smaller.
Negative mass is one of those attempts, and it's worth noting that they aren't clinging to the concept, they're simply suggesting that it's one possibility that can be tested. In other words, they actually are using Occam's Razor. In this realm, nothing is simple, which makes the Razor harder to use.
I'm thinking of the code-morphing, similar to Transmeta. From where I learned about it, the runtime translation target was called micro-ops. We have different definitions. Someone I once knew referred to micro-ops (my definition) as "caveman primitives."
Still, it's an internal CISC->RISC translation, and the retirement unit hides that when it's all done.
I suspect you're confusing micro-ops with microcode.
Current architectures (not all, but not just Intel) decompose the user-visible instruction set into a stream of micro-ops, (more primitive instructions) and send that stream to a dispatch unit. The dispatch unit resolves dependency issues and as requirements are met, sends the micro-ops to one of a series of execution units. As micro-ops complete, their results are sent to the retirement unit. Note that between dispatch and retirement, the architectural registers have effectively disappeared, and are reassigned at retirement.
Microcode is a completely different thing - usually the opcode is translated into a subroutine entry point, and a (typically) classic Harvard-style computing engine interprets the user-visible instruction set. But it's all in lock-step, not the controlled chaos of micro-ops.
Me too. I'd just rather postpone that day as much as possible, and have a good time while getting there.
On an achy day, my mother used to say, "Never grow old." However upon further consideration, I think growing old is usually preferable to failing to.
(Many caveats apply, "Growing old" is meant in the physical sense, of course making lifestyle choices to retain capacity. "Growing old" in the mental sense is also something of a choice.)
> * This guy is dead.
I get the distinct feeling that there are quite a few in the US govt and elsewhere that would like to help Snowden achive the same status.
Get hold of portable property. -- Charles Dickens, "Great Expectations"