42 comments

[ 0.25 ms ] story [ 47.6 ms ] thread
tl;dr Rant about complexity of https://en.wikipedia.org/wiki/Byzantine_fault_tolerance and it's practical uselessness.
I didn't know there was an area of CS called "Byzantine fault tolerance". Now, I'm completely astonished by both the intelectual honesty of people that call their field something that reads "known impossible problem that can only be dealed with by increasing complexity" and that the same people try hard to deal with the problem without bringing extra complexity.
Your tl;dr does this paper a serious disservice. It is at once more enlightening and more entertaining than most other papers.

It's also accurate.

Yes, but its a fitting one that does pretty much characterize the field.

It also highlights how remarkable the Bitcoin whitepaper is…

You realize that this is a humor column for USENIX's online magazine, right?
A different topic, but http://research.microsoft.com/en-us/people/mickens/theslowwi... is a good read too.
Quote from The Slow Winter

>> Unfortunately for John, the branches made a pact with Satan and quantum mechanics during a midnight screening of "Weekend at Bernie’s II." In exchange for their last remaining bits of entropy, the branches cast evil spells on future generations of processors. Those evil spells had names like "scaling induced voltage leaks" and "increasing levels of waste heat" and "Pauly Shore, who is only loosely connected to computer architecture, but who will continue to produce a new movie every three years until he sublimates into an empty bag of Cheetos and a pair of those running shoes that have individual toes and that make you look like you received a foot transplant from a Hobbit, Sasquatch, or an infertile Hobbit/Sasquatch hybrid." Once again, I digress. The point is that the branches, those vanquished foes from long ago, would have the last laugh.<<

This guy is going for the HST approach. Funny but opaque as well.

Another good one, titled "Mobile Computing Research Is a Hornet’s Nest of Deception and Chicanery" is at http://research.microsoft.com/en-us/people/mickens/nestofhor....
Wow, _somebody_ doesn't like windows phones

but seriously, that article had nothing to do with research. it seemed to be a rant at third-rate smart phones from 3 years ago.

Fits my experience with my Nexus 7 pretty well.

- The web-browsing experience is pretty bad (though web developers are as much to blame for this as the browser devs.) Mobile-optimised sites have text that's too large and advertising dialogs that forcibly redirect me to places I don't want to go. Buggy banner ads trying to stick to the bottom of the page bounce around the page as I scroll, take up too much of the viewport, and never dismiss when I hit the "X".

- The touchscreen does play up from time to time, sometimes when I'm using it and sometimes when I'm not. Two days ago my tablet opened Facebook and tried to send "monkey with typewriter" messages to old friends...

It doesn't make calls either, but I can forgive it for that... My heirarchy of needs begins with more "real computer" things, not "real phone" things. To type a short comment like this I'll first plug in a keyboard (which will cause all programs to crap themselves....) Anything more involved, though, and I'll consider spinning up an EC2 instance and doing it over SSH.

My favourite Mickens is The Night Watch: http://research.microsoft.com/en-us/people/mickens/thenightw...

"HCI people discover bugs by receiving a concerned email from their therapist. Systems people discover bugs by waking up and discovering that their first-born children are missing and “ETIMEDOUT ” has been written in blood on the wall."

and also

"I HAVE NO TOOLS BECAUSE I’VE DESTROYED MY TOOLS WITH MY TOOLS."

Just wondering: what does the [scribd] mean in the title? I thought this was a warning for the poor usability and difficulty to download from site http://www.scribd.com. This links directly to a pdf, so wouldn't a [pdf] tag be better? I've noticed this on a bunch of posts recently.
It's a HN auto-tag, to promote a Scribd, a YC company.
More specifically, it's actually a link to a copy of the same pdf hosted on scribd, the intention being to offer a means of viewing the pdf that's less cumbersome than Adobe Reader.
Interesting. I never clicked on it so I never noticed it was an additional link!

I've never liked the scribd interface, but what's more is that Chrome and FF have their own, decent, pdf viewers which are light-years better than either the scribd or adobe viewers.

I wonder how much HN gets to put that link there.

HN is owned by YCombinator. Scribd was founded under the YCombinator program. There's really not much to it.
I never see any other links like that for YC companies (and there are a lot of YC companies) on HN; that's why I was curious.

Aside: Why can I respond to you and not cmelbye?

> Why can I respond to you and not cmelbye?

There is a timer on new posts before anyone can reply. It's to force people to think a little before they post and is pretty effective in rate-limiting flamewars.

Not a scribd fan , but to be fair, they have been around longer than FF and chrome PDF viewers have.
What? You think Scribd pays their investor to promote their product? Isn't it much more likely that their investor is promoting the product they invested in for free?
I couldn't think of a way to make Scribd's user experience any less enjoyable. I usually give up after the 3rd page hangs and navigation and even reloading fail.
scribd may be the only means of viewing a pdf more cumbersome than Adobe Reader.
I always took it as a warning that "this will be a hassle to access".
(comment deleted)
The OP brings up a very old story in fault tolerance, reliable systems, and systems management: In practice a very significant fraction of all outages are due fairly directly to those famous words of the HAL 9000 "human error" or people in the server farm 'bridge' or the network operations center (NOC) doing the wrong things.
That was far funnier than it had any right to be.
CS background for people who haven't heard of Byzantine Fault Tolerance:

Original Lamport Paper - http://www.cs.cornell.edu/courses/cs614/2004sp/papers/lsp82....

'Practical' Byzantine Fault Tolerance -http://www.cs.cornell.edu/courses/cs614/2004sp/papers/lsp82....

This is the second article I have read written by this author and I've found both of them to be mildly informative and wildly entertaining. Sort of like a Dave Berry of computer science writing.
I agree, he is like Dave Berry. Absolutely horrible.
I was also getting a strong Dave Barry vibe. It's not bad - but it's easy to let the schtick get old, and Dave Barry is still alive. Maybe the guy is testing the theory that if you understand it well, computers are as crazy as Florida and thus capable of generating as many bizarre anecdotes suitable for column grist as Florida. Plausible.
I had absolutely no idea what a Byzantine fault was and to be honest I probably still have a very peripheral grasp of what the author is trying to make fun of. What I love about this is that you can sense the buildup of frustration that the author felt which then got released in the form of this paper, hilariously entertaining.
Very little real info and too much of what author probably believes to be a humor, but it's just pretentious BS...
Nah, the author thinks right, it's humor.
I chuckled. But, like it or not, lesser trust requirements are CRUCIAL in the immediate future of computing and therefore so is research on Byzantine Fault Tolerance.