Newer posts are loading.
You are at the newest post.
Click here to check if anything new just came in.
datenwolf
22:02
I'd like to comment on what was said about "mindset eras" by taking a tanget on rocket technology:

Mindset of the 1960-ies and 1970-ies:
Flys you to the Moon – 'nuff said

Mindset of the 1980-ies:
Builds you an overcomplicated orbiter with several fundamental flaws emerging from it's top-down construction. The Challenger got destroyed in the end due to problems in maintainung the SRBs, making them suceptible for blowby. The Columbia's fate was sealed by the delicateness of the thermal insulation system. The SSMEs had their wealth of problems too, leaving the NASA engineers wondering why never anything bad happened from them.

Mindset of the 1990-ies:
"We can manage the 1980-ies technology by implementing the right protocols." → No significant development in space technology were made at all, because everyone was busy fixing stuff. Meanwhile the russians Soyuz rocket from the mid 1960-ies became the workhorse of manned space flight.

Mindset of 2000-2010:
"Fuck this shit, back to the roots." Just look at the NASA Constellation Program. The Ares rocket uses the J-2 rocket engine – the engine of the Saturn-V, i.e. 1970-ies technology.


The era from 1950 to the 1970-ies is in my opinion the era of the best engineering so far. And whenever I watch "The Right Stuff" or "Apollo 13" I get this feeling of "I wish I could work in that era with those guys."

About that remark, that Linux lacks the revoke() system call and thus ConsoleKit can't do it's job properly: That's exactly that kind of mindset that got 14 people killed (in the Space Shuttle program). If you know that a targeted OS won't have that feature anytime soon, don't design security critical programs assuming it was there. Also revoke() is kind of trying to push the toothpaste back into the tube. And ConsoleKit is intended for doing fast user switching – frankly I don't understand how that's to work properly with revoke(), when everytime you switch the user, a set of programs get their file descriptors invalidated. That will require to add further code in several programs provisioning that, so that they don't crash or deadlock. ...which will require additional logic for dealing with the consequences of that code and so forth.
The solution lies in the kernel indeed, but not by some magic syscall. I'm thinking along the lines of containers and namespacing.

This is what I (and many other people) so totally dislike about the contemporary Linux desktop design. I say "Linux" because that's what they are targeted for.

Now I may be totally wrong about that, but: I bet a truck of Mate, that our software will be much simpler and similar to the 1970-ies code than to what we've got today in at most 20 years from now (that's the age of Linux this year).

It's easy to say "told you so", but when it got introduced, I forsaw the fate of HAL from the very first day. And as devfs got replaced by udev I told from the very beginning there should be a merger between devfs and tmpfs. 6 years later we got exactly that.

I have those same hunches about Wayland (although it's too early with that one to be conclusive), those *Kits and numerous other stuff.

Now I decided on the following resolution:
I'll let Freedesktop go it's way, but I'll not follow it anymore, I'll go my own way instead. If that means a lot of customary coding work, so be it. If Freedesktop's advantages influence the Linux Kernel, I'll just won't configure into my system.

If my way fails, then it's only my problem, but you won't hear me complaining about my stuff. If however the whole Freedesktop thing fails, well.

In the 27c3 talk I could easily have Lennart removed. But that would have been a very, very bad move. I've just read the IRC backlog. Makes some interesting reading if you filter out all that popcorn. A lot of people were wondering, what those "*Kits" are, seemed even unaware of their presence on their systems. If the only thing that whole mess did was making people just take a look into those, then I succeeded.

Rest assured: I was never anything near to breaking in – easyly said, but true. Yes, I was gaining some facial colour, who wouldn't have? Did Lennart have some valid points. Yes, nothing in denying that. But still I stand to my general opinion on the state of Linux desktops.

Which leaves me to ask all fellow coders and hackers to flea (= give it a line-by-line audit) the code of D-Bus, ConsoleKit, PolicyKit, RealtimeKit, and systemd. And even more important: Design and implement alternatives, to add evolutionary competition. That Freedesktop monoculture does no good.
Reposted byg33ky g33ky

Don't be the product, buy the product!

Schweinderl