« first day (3036 days earlier)      last day (1974 days later) » 

Bob
12:38 AM
 
Dropping back to share this:
 
Bob
1:12 AM
> The Bureau of Meteorology has updated its severe thunderstorm warning, saying the severe thunderstorms which have already brought a month's worth of rain could also bring hail.
welpers.
houses are flooded O_O
stations are flooded
weeeeeeeeeeeeeelp
The torrential rain has hit Lewisham Station hard this morning, flooding the entire subway of the station. #9News
 
Bob
1:58 AM
so, that ticwatch in a too-big box?
here's the charger:
that's ... a lot of wasted space
 
2:45 AM
One box fits all!
 
2:59 AM
> Moore’s Law has come to a grinding halt
 
HAMMERZEIT!
 
@ChatBotJohnCavil rofl
Now what?
 
 
4 hours later…
Bob
6:31 AM
O_O
the way the clouds are spinning on radar doesn't look nice at all
and apparently we were supposed to go home an hour and a half ago
....weeeeeeeeeeeeeeeeeeeeeeelp
 
 
1 hour later…
7:46 AM
Good morning.
Been a while.
How is everyone?
 
8:04 AM
Because I was bored. Freenom Order Number 2912570354.
Mapped rootaccess.freedomains to this room, but I don't think SE chat likes iframe, it keeps showing "chat.stackexchange.com refused to connect".
 
Hmm. XSS ....
Maybe hsts.
 
9:03 AM
morning
 
I've heard the USB 2.0 protocol described as both half-duplex broadcast, as well as unidirectional multicast. Is one of the descriptions more accurate than the other? I would think the former is more accurate because the communication is not truly unidirectional and can send data in both directions, only one at a time.
 
@forest sadly the dude who can totally answer your question dosen't hang out on chat
 
Oh well.
 
@JourneymanGeek who?
 
(not even joking. There's a specific user who trawls for USB questions and posts insanely detailed, extremely technical answers)
@Burgi I think he went by Ali Chen
or she. I don't even know
 
9:09 AM
Oh neat.
 
i haven't seen them
 
Then more generally, for a multidirectional protocol (whether half-duplex or full-duplex), how would one describe it if one direction is multicast and the other direction is not?
 
magic?
 
well usb is point to point, so it'd never be multicast, or broadcast.
i guess if you're throwing in a hub it could become so
 
It is multicast in the master->slave direction for 2.0 and older.
 
9:15 AM
I thought it..... didn't do that
tho tbh I've never really checked
 
Nope, it does. Makes it possible to spy on data being written to an external USB hard drive just by plugging in a device that can sniff raw USB traffic into another port on the same internal hub. Doesn't work for 3.0 though, since xHCI routing involves switching.
 
@djsmiley2k there's often internal hubs ._.
 
21
A: Can a connected USB device read all data from the USB bus?

AdiMost likely yes, but it depends Much like PATA, SCSI, and Ethernet devices, USB devices don't directly connect to the computer. They connect to a Host Controller that manages all signaling and communication. All ports are connected to something called a Root Hub, and to each Root Hub you may con...

@JourneymanGeek If your system has more than one port, it has an internal hub.
Actually even if it has only one, it has a hub, just with only one output.
 
9:47 AM
@forest yup
 
The USB protocol is fun. I tried fuzzing it once but nothing much happened.
EHCI and xHCI seem pretty well designed.
Then again, I only did rather slow, inefficient fuzzing through hardware...
I imagine a comprehensive protocol-aware fuzzer could find something.
 
hmmm
/me bears it in mind
 
10:18 AM
Hey, I fixed it! Once I changed frame (cloaking) to HTTP 301 forwarding, I got it to work.
You can now get to this chat using rootaccess.tk
 
why would i?
 
if you're on another system and ... actually, I have no idea.
 
it timed out here
 
HTTPS doesn't work, only HTTP.
 
@forest Oh, ok! Got it. Seems https is a paid thing.
 
10:23 AM
this seems like the prelude to a MitM attack
 
you mean one day, you open that open this page through that link and everything looks weird.
Sure, eventually. I'm too lazy for that though.
 
@Burgi Indeed it is. All it'd take is an injected redirect to a version of SE that uses an IDN homograph attack to prevent you from noticing it's a different site.
 
The internationalized domain name (IDN) homograph attack is a way a malicious party may deceive computer users about what remote system they are communicating with, by exploiting the fact that many different characters look alike (i.e., they are homographs, hence the term for the attack, although technically homoglyph is the more accurate term for different characters that look alike). For example, a regular user of example.com may be lured to click a link where the Latin character "a" is replaced with the Cyrillic character "а". This kind of spoofing attack is also known as script spoofing. Unicode...
#TIL #blackhat stuff
 
@Burgi in general, you should never trust .cf, .tk, .ml, or any other freenom domains
 
@rahuldottech yup
 
10:29 AM
Jun 21 at 21:22, by rahuldottech
@Ave I'd been hearing about freenom hijacking their free domains for a long time, but it had never happened to me, in the many years that I'd been using them. Then like a couple of days ago it finally happened. Luckily, I'd moved all my important stuff to proper domains owned by me
7 mins ago, by Nick
@forest Oh, ok! Got it. Seems https is a paid thing.
 
 
The only people who should pay for HTTPS are organizations like banks, and those who don't have the skills to set up a free cert, because it is ridiculous easy to do so at this point.
 
 
@Nick why tho
 
@rahuldottech it was free, and so was I.
 
10:34 AM
@rahuldottech No one should ever pay for a certificate, only for EV.
 
10:46 AM
@forest no one needs to ;p though occationally apparently you don't get everything you need from LE
For personal use? hell yeah
 
What's wrong with getting everything from LE? They're the best out there if you don't need EV. I think they even support wildcard certs now.
 
@forest iirc SE looked at them, and at the time they didn't do wildcards
and I guess at times there's some obscure feature you need they may not support
 
If regular old DV is enough, then what other features could there be that needs the CA itself to support it? I mean, maybe if you need a strangely-sized key or one which uses non-standard algorithms...
 
idk. Its what I use myself ;p
 
Same. I used WoSign^WStartCom before that.
 
10:51 AM
I didn't have https before that
 
I've not set it up yet, because I don't really host anything anymore :(
 
quite honestly it was a "eh, its so rediculously simple I'd be an idiot not to use it" thing
actually wasn't since I was using lighttpd
now, with ngnix...
 
You can still get a cert very easily even if you have an unsupported httpd.
 
you can, but its a little extra work
with ngnix, its pretty much a no brainer
 
I used to use nginx, but their security track record is sad. :(
They remove as many error checks and asserts as they can for performance.
 
11:34 AM
This is interesting.
 
12:01 PM
Is this all still up to date?
 
12:49 PM
@forest I'd just go with bzip and gz
I don't know if it's still 'true', but it's pretty damning and I don't have the time to research it.
 
1:15 PM
It looks like it is still true. For some reason I thought it said 2004, not 2014.
 
I can't wait for mine to come.
 
1:38 PM
What switches you gone for ?
 
Bob
 
Hmm. Annapurna Labs -- the guys who made the SoC in my router that was buggy as fuck for the first 6 months of owning it, but has really settled in thanks to Netgear patching
 
@djsmiley2k Brown.
 
@djsmiley2k Some of the claims which provide data from tests has to be trusted unless you try to do the tests yourself, but some of the claims involving the decreased error detection accuracy when using SHA-256 or CRC64 when compared to CRC32 (something that's normally the other way around) is correct, given my admittedly limited understanding of cryptography and burst error detection (I've read the Koopman paper on CRC polynomial choice, but that's about it).
 
an a1.medium spot instance is $0.0084 per hour, or $6.048 per month, for a server with 2 gigs of RAM running on one ARM core
$96.69/month for 16 ARM cores and 32 gigs of RAM
 
1:53 PM
@HornOKPlease though... was the SOC buggy or the firmware?
consumer routers are kinda famous for... well...
one of the nice things with a general purpose machine is if an OS or distro sucks, there's an option
 
Today is the first day that modern signaling tech is live on the entire 7 line. After we finish optimizing and automating the system and performing other signal work over the coming weeks, we’ll be able to run more trains, more reliably. Thank you for your patience.
 
not always options with many consumer routers
 
Main St-bound 7 train service is delayed because of a network communication issue.
Less than one hour after CBTC came online.
 
@JourneymanGeek a lot of OpenWRT/DD-WRT hackers were saying the SoC was horribly buggy, but Netgear seems to have tamed it, probably with CPU microcode updates from Annapurna/Amazon, plus their own firmware work
 
@HornOKPlease openwrt I trust a little...
DD-WRT is...
@HornOKPlease eh, I guess.
 
1:59 PM
I wonder if Annapurna contributed code to upstream Linux for their SoC? there's a hideous amount of ARM code in the kernel now
from umpteen different ARM-based processor silicon shops
 
@HornOKPlease likely
 
hm. Netgear could've just updated to the latest Linux kernel then :P
 
@HornOKPlease that's the slightly annoying thing about ARM :p
 
7 and 7 express trains are running with delays in both directions because of a network communication problem at Queensboro Plaza.
 
Its really good if you want to buy an IP block
 
2:00 PM
@bwDraco ah, I see the 7 line is having a Fallout 76 launch
 
but no one really seems interested in the platform
 
"it's installed, but it won't run"
 
I mean, they talk about it...
but I can buy an arm power workstation...
the closest things I can think of is that wierd 99 dollar laptop, and Bunny Huang's Novena...
which is a neat bit of kit, admittedly...
 
@JourneymanGeek it's super fragmented, that's the worst thing, because supporting all popular x86 processors involves two vendors' microcode (Intel and AMD) and two vendors' processor-specific kernel code (the same). And each has a massive install base that guarantees you won't have, like, low-level CPU issues with Linux.
 
@HornOKPlease Just 45 minutes after they declared CBTC was online, the train gets delayed. facepalm
 
2:02 PM
@HornOKPlease techically 3
via keeps shambling on.
 
oh hah
 
apparently there's a chinese company that licenced it from them (idk how...) and makes evolved via chips
 
@JourneymanGeek I've heard virtually nothing about Via these last few years.
 
does Via have to license from both Intel and AMD?
 
@HornOKPlease I suspect just AMD
the patents for oldschool x86 are dead
 
2:03 PM
but Intel owns the x86 32-bit ISA
ah
 
@HornOKPlease its probably old enough to drink in most places at this point
 
@JourneymanGeek that seems really shady, I wouldn't be surprised if they stole/copied Intel stuff
 
@HornOKPlease naw, wouldn't need to
They basically went "Hey, Via! We'll pay you a bucket of money for you to design a chip and licence it to us..."
and via's like "Oh ok!"
Via has been developing x86 chips the entire time
 
but I don't think Via is good enough to produce a chip that can rival Kaby Lake??
 
2:06 PM
> Chinese government
 
quite honestly, I'd not be surprised to see them in machines like my router ;p
 
Embedded chips, IIRC.
 
Tencent, also sponsored by the Chinese government, has an entire department dedicated to stealing foreign IP
 
@HornOKPlease the idea is to have a non US controlled source of x86 processors I think
@HornOKPlease wouldn't even really be needed here tho
Via (who amusingly, are taiwanese) probably have all the IP they need
 
@JourneymanGeek why not? they may have a license for the ISA, but actually building a performant processor requires a ridiculous amount of investment
 
2:08 PM
@HornOKPlease I doubt it would be anywhere as efficient as a modern intel or amd chip
 
the IP from AMD/Intel wouldn't contain the techniques for optimization and performance and such
 
but it would keep them in the game long enough
and if memory serves - via built slow but extremely power efficient chips
(with their own quirks... remember that via c3 'hack'?)
 
IIRC the focus of VIA has always been on extreme cost effectiveness, not any particular quality, performance or efficiency goal
 
@HornOKPlease maybe? ;p
 
read: what's the cheapest thing we can throw out there that technically runs
their top-end graphics card is like $50
and that was when it was brand new, a couple years ago
@forest fascinating
seems pretty biased since it's written by the author of Lzip, but then again, Lzip works so he probably knows a thing or two about compression formats... and the actual meat of some of what he's saying seems true to my limited knowledge of low level file formats
 
2:17 PM
@HornOKPlease so they take... 4 of those, stick em on a die...
 
IIRC Via QuadCore was a two-die MCM.
 
2:30 PM
(4C Isaiah, 4 MB L3$)
 
@rahuldottech Did you mean
user image
3
 
Not too sure where their 28nm chips went...
 
Bob
@HornOKPlease Scaleway is cheaper for the same thing and doesn't charge out the arse for data transfer and storage. But less reliable :P
 
@HornOKPlease Well he does give a disclaimer, but it seems technically sound, assuming his understanding of the xz format is correct (which looks very likely).
 
For whatever reason, VIA is still doing 40nm...
 
Bob
2:37 PM
@HornOKPlease "The potential". It's probably 95% bullshit. Let's start with how they seem to be comparing 8 cores to 4...
And, yea, performance reports from the manufacturer. *Dave sniff* I smell bullshit.
Sounds like when HiSilicon/Kirin was bragging.
 
@Bob the bigger point I was trying to make was...
THEY EXIST
 
Bob
Or when MediaTek/Helio was bragging.
 
Anyhoo, driving lesson in 20 minutes.
 
@Bob Dave sniff :D
 
@Bob The point of having 10 cores is having as many as three groups of cores to choose from for power efficiency or performance. Remember that power usage goes up disproportionately as you increase performance, whether it's IPC or clock frequency, and a complex core will often consume more power than a simpler core, even at a lower frequency matching the performance of a higher-clocked simple core.
 
2:41 PM
@bwDraco also...
its a bigger number
 
@Bob ah, I missed that... heh... then even if it were true that their CPUs benchmark comparably to a Kaby Lake i5 (I'm guessing U series, but they could surprise with desktop perf), that means they're getting only half of the oomph per core (IPC + clock speed combined) than Intel
AMD is much closer to Intel on that front currently
 
@HornOKPlease with would be unsurprising
 
That's why, for example, nobody uses low-frequency Cortex-A57 cores.
 
as far as I know, the fastest ARM core currently (save, perhaps, those crazy server chips that can consume way more power) is in the 2018 iPad :P
 
Apple's cores are designed for maximum single-threaded performance over core count.
 
2:44 PM
which may still have more total compute power when you include what its GPU can do, with CPU+GPU going full tilt
 
They have more engineering resources than even Qualcomm, and can tightly couple hardware and software.
 
Bob
@HornOKPlease I feel like as soon as you're talking about long-term archiving, especially robustness against corruption, talking about recovery from bit-flips in compressed data is going down the wrong path.
More copies!
 
The only company other than Apple with anywhere near that kind of R&D budget for an application processor is Intel.
 
Bob
@JourneymanGeek Mostly the bigger number.
big.LITTLE makes sense, to a point
 
like old school clock speeds, or blades in razors...
@Bob it does
 
Bob
2:47 PM
But beyond that point the increasing complexity just opens up more room for error.
Also. Chinese companies like big numbers.
 
@Bob I read it as more about consistently being able to detect bit-flips, not recover from them -- and then be able to know where the bit flips are and maybe be able to salvage some of the data if not all of it, by discarding a hunk or something
 
Bob
@HornOKPlease ...salvaging data = recovery
 
and also about file format documentation so, when in 50 years GitHub is gone and nobody has source code to an Xz decompressor anymore, the file format is consistent enough that someone can write a decompressor from scratch to recover data and it'll work on all files of the format
 
Bob
IMO that's not the job of the archive/container format.
That's the job of the underlying storage.
@HornOKPlease Yea, having the format documented is important.
I can agree on that.
*shakes fist at Mozilla*
 
@Bob he seems to think it's the job of both, like defense in depth... and also, since Xz is basically like .mkv for lossless compression, where it can support umpteen algorithms for hashing and umpteen algorithms for compression, it's harder to just "write an Xz decompressor" that will work on all Xz files
 
Bob
2:50 PM
I had a lot of fun trying to figure out how Mozilla's lz4 variant worked...
 
and then he said the file format doesn't appear to be versioned properly, showing the example of the file tool not telling you what version an Xz archive is
and then there are easy ways to corrupt the data by having the compressed data block end in a CRC
lots of little problems that will only rarely come up, but make it a needlessly flawed format
 
Bob
@HornOKPlease lol.
> On Unix-like systems, where a tool is supposed to do one thing and do it well
 
@HornOKPlease Unfortunately some of it is undocumented, which he pointed out.
 
Bob
@HornOKPlease I agree with the issues around looseness of the spec.
I don't agree that it's the archive's job to detect bitflips.
 
@Bob Ugh, I had to do the same thing. :/
 
Bob
2:53 PM
Thing is, compression formats in general have really gnarly documentation.
That goes for image compression too.
My brain still hurts from trying to make sense of PNG.
I'm not even gonna try to understand video compression.
 
Compression is far simpler than something like PNG or video compression.
In part because each individual I-frame itself often uses, say, DEFLATE.
 
Video seems to be the worst, because it's almost like each compression/decompression codec and video format handler has its own special interpretation of the spec
 
And that's not to mention things like motion prediction, P-frames and B-frames, macroblocks, motion compensation, and of course audio...
 
dozens of little flags (some documented, some not) about codec "options"
the chance of a random file being editable in Adobe, Quicktime and Handbrake and playable with VLC, Gstreamer and Firefox/Chrome is pretty low unless you stick to a few, extremely well defined standards
 
That I actually find amusing. x264 has a tuning option called "touhou", which is a reference to a rather obscure Japanese video game (obscure in the west) which has a cult fan following. It was added specifically to optimize replay videos from that game.
 
2:58 PM
!!/wiki Touhou Project
 
The Touhou Project (Japanese: 東方Project, Hepburn: Tōhō Purojekuto, lit. Eastern Project), also known as Toho Project or Project Shrine Maiden, is a series of Japanese bullet hell shooter video games developed by the single-person Team Shanghai Alice. Team Shanghai Alice's sole member, ZUN, independently produces the games' graphics, music, and programming.Plots in the Touhou Project games revolve around the strange phenomena occurring in Gensokyo, a fictional realm inhabited by humans and yōkai, supernatural beings. Prior to the events of the games, Gensokyo was sealed off from the outside world...
 
Bob
@HornOKPlease Oh god. The "little flags" in ffmpeg.
 
Bob
@forest Lol. Tohou is obscure?
 
It's a shoot-em-up game, basically like that, but often far harder.
@Bob In the west? 99% of people have no idea what it is.
 
Bob
3:00 PM
@forest It was pretty darn popular in the flash game era... popularised bullet hell/heaven, iirc :P
 
Popular among diehard danmaku fans and weaboos, but not much else.
 
Bob
Though that might've been the Raiden series
 
Maybe it is getting more popular though since it was put on Steam.
But yeah, video formats are insane. It's like an entire operating system in complexity.
 
@forest The 16th mainline game did reach Steam...
 
Dammit if it actually has become popular then that means I'm going to have to hate it. :(
 
Bob
3:02 PM
lol
 
Oh my would you look at the time. I gotta get going.
ttyl
 
See you in a bit.
 
 
1 hour later…
Bob
4:07 PM
Hm.
Tempted to buy a 500 GB Samsung T5... 150 AUD
 
 
1 hour later…
5:11 PM
 
!!wiki karl popper
 
Sir Karl Raimund Popper (28 July 1902 – 17 September 1994) was an Austrian-British philosopher and professor.Generally regarded as one of the 20th century's greatest philosophers of science, Popper is known for his rejection of the classical inductivist views on the scientific method in favour of empirical falsification. A theory in the empirical sciences can never be proven, but it can be falsified, meaning that it can and should be scrutinised by decisive experiments. Popper is also known for his opposition to the classical justificationist account of knowledge, which he replaced with critical...
 
@XKCD too obscure for me
 
5:32 PM
Most of the time when I get an email recalled, it leaves my local copy that I've already read. However, an admin just recalled an email and it removed it from my inbox. Is this a setting in outlook or Exchange server(?), or have I just been doing it wrong? Lifehack: I actually added a 3 minute delay on my outbox to avoid the recall panic (oops! where's the recall option? Google? Anybody!?)
Since adding the delay, I am nearly 100% satisfied with every email I send out...
 
 
2 hours later…
7:41 PM
...welp. On most HP laptops that use barrel charge connectors (whether 7.4mm as used on older systems or 4.5mm on current machines), there's an ID pin in addition to the power contact, used to determine, among other things, power adapter capacity. If you insert the connector too slowly or don't push it all the way in, it'll power the laptop but it won't charge the battery because it can't identify the charger.
On some systems, the performance will be reduced if it can't identify the adapter in use (especially if the system requires a higher capacity unit, e.g. 65W or 90W). On this ENVY x360, it will still allow full processor power, up to 20W sustained and 30W peak.
That's because strictly speaking, it only requires a 45W adapter, and that's the smallest HP makes with the 4.5mm barrel connector.
(though it will charge faster under load if given a 65W adapter)
Retail HP Smart AC Adapters use the 4.5mm connector, with a 7.4mm adapter supplied for use on older systems.
I'm actually exploiting this behavior to prevent the laptop from charging when powered on, so that I can keep the battery to a more optimal state of charge when plugged in for extended periods of time (constant 100% is not ideal for long-term service life). The system will still charge the battery when powered off or in sleep. It'll reset to normal if unplugged and plugged back properly.
(note that my first laptop, a 15.6" HP Pavilion dv6z-3000, would do the same thing, but that system required a 90W adapter, and would actually pop up a warning saying that the AC adapter capacity is too low if it fails to identify the AC adapter because I didn't plug it in all the way; I get no such warning on Stolas)
Note that this system can also accept an HP-branded (not USB-PD) USB Type-C charger.
 
 
2 hours later…
11:04 PM
not bad for rendering Elder Scrolls Online with 8 characters on the screen in high detail, HBAO, AA, and 85-90 FPS
PerfCap Reason: Idle :)
 

« first day (3036 days earlier)      last day (1974 days later) »