FOSDEM 2016: more boring shit
Let's take a look at my annotated copy of the FOSDEM 2016 main talk schedule, shall we?
Probably a lie.
Places we're NOT taking it include "Interoperability City," "Lake Reliability," and "The People's Republic of Sanity".
Alternate subtitle: "we didn't give a shit about any of these people before we saw them on the news, but look what we can accomplish when we need to feel righteous about our shitty hobbies"
I think the printer's jammed, we need you guys back at the office
You know that huge pile of bad software the "devops" people wrote so that they wouldn't have to ever actually install their software? This guy wants to make that the norm. Everywhere.
The speaker confuses reproducible builds with package signing, then makes a ton of excuses about why his talk summary is meaningless noise.
Except it's not really "a linux session" as much as it is "onboard cpanel shit for people who should not be trusted with computers." Naturally, the strongest evidence that this is a mistake is its inclusion in Fedora by default.
A GNOME developer talking about security.
This isn't "open source" in the "use this how you want" sense. This is the "help us do work for no money" sense.
Every goddamn thing, to the author's surprise. Formalization of the process doesn't seem to improve anything, it's demonstrably irrelevant which formalized process your team adopts, and the author is "trying to make code review work as expected with my research," which will probably be too dull to be hilarious.
Setting aside the fact that "machine learning" has yet to prove worthwhile in basically any current implementation, and the fact that "big data" is almost always code for "shitty programmers who ran out of memory," this abhorrent pile of dogshit is intended to strap together a bunch of computers scalemp-style in an attempt to make single-threaded hobby projects relevant. Absolutely disgusting.
Everyone's favorite also-ran clustered filesystem continues thrashing around in irrelevancy, but at least they're spending a lot of time rewriting existing functionality.
This one came as something of a surprise, since I've been using linux with multi-petabyte filesystems for many years. Unfortunately, the actual speech seems to be based around a version of Plato's allegory of the cave, but instead of shadows on the wall the speaker sees webshit technology and assumes there is no other reality. Because webshit runs in userspace, POSIX is irrelevant and filesystem i/o should be moved out of the kernel. Yes, the author works on Gluster.
The speaker assumes that since FPGAs are getting cheap, all the problems will go away. The fact that all available FPGA development kits are either ruinously expensive or iredeemably shitty does not appear to factor into this.
Another entry in the "pretentious three-part FOSDEM title" history books. Seems like it could be an interesting talk if it were just about FM radio and less about GNU/crap.
"so much industry momentum" that nobody gives a shit at all! Incedentally, if you've noticed how bad your haswell video support is on linux, this speaker is part of the reason why. Cheers!
Seems like a risky move putting the entire one-hour presentation directly into the talk title, but maybe it'll work out. We'll have to spend less time getting excited that they've finally got around to supporting laptops made a mere six years ago, or that post-Ivy Bridge laptops can never be supported. At least you can install this on your servers, thereby destroying all the useful remote operations shit that might have otherwise made your life easier!
The answer to this has always been "yes, as long as you're willing to sign an NDA before getting audit access to the code." The talk, predictably, has nothing to do with this question, instead focusing on handwaving an open-source software that everybody already uses.
Bonus answer: why the fuck would anyone pay for the infrastructure under these circumstances
Facebook figured out InnoDB sucks. Unfortunately, they're so rich that writing another database backend seemed like a good idea. Oh well.
No, the speaker has never designed a kernel API. He just writes books about them, and asks for money in exchange for "training" about them. I really don't understand why there is a talk here aimed at kernel developers, since they already have a mailing list, and their own conferences, and firmly-entrenched opinions on the topic.
some webshit thinks you can solve social problems with git
Ah, this must be the sort of "momentum" that Vulkan experiences. In practice, an ODF document means one thing: the user downloaded OpenOffice and forgot to save in Word format. Good news, though! A quick email should be enough to get them to re-save it and send it to you.
In which it is revealed that their fake version of Office365 is wildly resource-hungry and subject to crippling security problems. At the end of the presentation, everyone who is surprised by this information will be given a Mac and a job at Facebook.
The Don Quixote of FOSDEM 2016. This speaker is attempting to speak to conference attendees about one of the lost arts of the internet: email that is not hosted by Google. He tries to make it interesting for the typical FOSDEM attendee by tossing terms like "http" and "json" into his slide deck, but then he ruins it all with technical discussion of hash algorithms and abstract syntax trees. If he just renamed the project "spam.ly" and gave the talk wearing jeans and a sport coat, Paul Graham would give him eighteen million dollars.
Last year I made fun of the systemtap team for their desperate attempts to seem relevant. This year they're back, and it's even worse. In 2015 there was a linux kernel vulnerability that had to do with a specific variable being signed instead of unsigned. Our team patched the kernel and redeployed and we were fine within hours. Red Hat's official "Band-aid" for several weeks was to use systemtap to bitwise-AND the value with INT_MAX, during runtime, on every boot of every computer. I tell this story to systems administrators as an example of how shitty Red Hat security responses are. Now they are giving a talk about how wonderful it is.
The second theoretical talk at FOSDEM: the speaker does not work in the telco field. Presumably this talk is a sales pitch for all those telecom executives at FOSDEM. Tons of those.
An introductory howto for people who actually want to know how to use the features their operating systems provide. Presumably scheduled after lunch on Sunday to give the engorged web crowd somewhere to sleep. I suppose it's possible that this talk is presented as a curiosity, but then it would have been titled "What people do in a real conference" or something.
In this talk, the speaker brags about Kubernetes fault tolerance by unplugging several shitty computers from a cluster of shitty computers. It is not mentioned that Kubernetes is sufficient for this task: it will generate its own faults regardless of what hardware you run it on.
Amit Shah finally gets around to explaining why QEMU live migration used to work fine and now it completely sucks.