|
Clearly they should be using an unsigned int
|
# ? Jan 1, 2022 22:41 |
|
|
# ? May 14, 2024 02:22 |
Beef posted:Clearly they should be using an unsigned int but what if they had to represent a BC date???
|
|
# ? Jan 1, 2022 22:42 |
|
Of course it is anti virus
|
# ? Jan 1, 2022 23:06 |
|
Jazerus posted:but what if they had to represent a BC date??? I tested it with today's date and it works, I don't see what the problem is. Can you point me to a specific customer that has that problem? Closing ticket.
|
# ? Jan 2, 2022 00:35 |
What? We need a fully compatible solution. We should count picoseconds from the big bang and until the most optimistic estimate for how old the universe will be. Anything less and we'll live to regret it.
|
|
# ? Jan 3, 2022 19:56 |
|
A 128bit unsigned int would store that pretty easily.
|
# ? Jan 4, 2022 04:30 |
|
Joda: picoseconds ought to be good enough for anyone Particle and astrophysicists: *cries*
|
# ? Jan 4, 2022 04:53 |
|
Foxfire_ posted:Joda: picoseconds ought to be good enough for anyone Have fun calibrating an absolute timestamp to anything tighter. Though if we follow timespec's lead we'd probably have a 64 bit seconds field, a nanoseconds field, and an attoseconds field.
|
# ? Jan 4, 2022 06:14 |
|
Dylan16807 posted:Have fun calibrating an absolute timestamp to anything tighter. Planck time units or bust.
|
# ? Jan 4, 2022 07:46 |
|
Can’t wait for processors to be advertised by planck lengths.
|
# ? Jan 4, 2022 14:55 |
|
At some point the distances relative to the amount of time measured just smear everything into a meaningless blur. Light travels about one foot in a nanosecond so at the datacenter scale, nanosecond level measurements if actually accurate turn into godawful physics homework problems. Light travels about 186 miles in a millisecond so millisecond level measurements are getting blurry between datacenters or between a datacenter and a client even if both have "the same" clock as observed from a non-moving frame of reference exactly between them.
|
# ? Jan 4, 2022 23:17 |
|
1337JiveTurkey posted:At some point the distances relative to the amount of time measured just smear everything into a meaningless blur. Light travels about one foot in a nanosecond so at the datacenter scale, nanosecond level measurements if actually accurate turn into godawful physics homework problems. You’re like half way to reinventing distributed systems with this post, and I say that with admiration. Specifically, “Time, Clocks and the Ordering of Events in a Distributed System,” which was inspired by special relativity.
|
# ? Jan 5, 2022 00:15 |
|
I had to write a basic distributed system at university and yeah it was frustrating to get stuff to execute in the order it happened across multiple physical machines, this was pre VM don’t guess my age. From memory I ended up with some super slow and bloated insanity that used one node in the distributed system as the primary and every other node had to ask it for sequence number for its transactions which each node would then wait until it got the number back before adding it to a queue for processing in order. If a number was missing in the sequence it would go and ask the primary to resend it. Think it was written in C and just forked processes off. It “worked” and I passed. At least it wasn’t the write a Java compiler in Java class that was taught over one week with a guest lecturer.
|
# ? Jan 5, 2022 00:32 |
|
lifg posted:You’re like half way to reinventing distributed systems with this post, and I say that with admiration. Specifically, “Time, Clocks and the Ordering of Events in a Distributed System,” which was inspired by special relativity. That's probably because I've read some of the work that's built on that regarding partial ordering of events but never got around to that paper. (Available here for anyone curious) Intuitively having a distributed clock never seemed worth the trouble versus simply accepting the partially ordered nature of things but it looks worth a read. Thanks!
|
# ? Jan 5, 2022 18:48 |
|
1337JiveTurkey posted:That's probably because I've read some of the work that's built on that regarding partial ordering of events but never got around to that paper. (Available here for anyone curious) Intuitively having a distributed clock never seemed worth the trouble versus simply accepting the partially ordered nature of things but it looks worth a read. Thanks! Generally speaking it tends to be a "If you have this problem, you're probably doing it wrong". Parallelize what can be parallelized easily (Ie lots of pure tasks that dont have horizontal dependencies). And do serially everything else. Unless you absolutely have to, in which case , its good to have a prescription for anti anxiety drugs, or thorazine if you've been doing this poo poo too long and still keep falling into the same hole.
|
# ? Jan 6, 2022 07:07 |
|
lifg posted:You’re like half way to reinventing distributed systems with this post, and I say that with admiration. Specifically, “Time, Clocks and the Ordering of Events in a Distributed System,” which was inspired by special relativity. It's also halfway to reinventing GPS
|
# ? Jan 6, 2022 18:04 |
|
duck monster posted:Generally speaking it tends to be a "If you have this problem, you're probably doing it wrong". Such a widely applicable statement. My favorite to throw it at would probably be: "We can't do X because it'll require too much live-ops work (read: manual data entry) every day." I'm remembering that thing about how SQL Server or something will crash if it receives something timestamped from the future, so it just adds 1000ms to everything to account for clock differences.
|
# ? Jan 7, 2022 01:51 |
|
Ranzear posted:I'm remembering that thing about how SQL Server or something will crash if it receives something timestamped from the future, so it just adds 1000ms to everything to account for clock differences. Ughh. Some of our application servers had trouble with NTP sync for a while, so we would see occasional 403s because the SSO server would produce JWTs that looked future-dated to the backend API. Telling ASP.NET to skip validating the token date was an obvious no-no, but giving it just a small clock tolerance would have been tricky, so we simply had the front-end wait 1000ms and then retry in case it received a 403 from the initial API call. The NTP issue has since been fixed but I bet the workaround is still there.
|
# ? Jan 7, 2022 19:05 |
|
NihilCredo posted:Ughh. Some of our application servers had trouble with NTP sync for a while, so we would see occasional 403s because the SSO server would produce JWTs that looked future-dated to the backend API. Hardly the worst thing... unless it can retry more than once.
|
# ? Jan 7, 2022 19:21 |
|
What is the reason for nbf anyways? Are there scenarios where auth servers issue future tokens on purpose?
|
# ? Jan 7, 2022 19:26 |
|
If you get banned from the system it's to prevent working around it by traveling back in time.
|
# ? Jan 7, 2022 19:49 |
|
Now, class, when having a simple utility written in bash, which takes one or more filenames as arguments, why is it a bad idea to have a built-in help that goes like this:code:
Edit: Yes, this mistake was copy-pasted to no poo poo literally 17 files. That I've found so far. bolind fucked around with this message at 14:01 on Jan 12, 2022 |
# ? Jan 12, 2022 13:02 |
|
You should be using underscores and not dashes in your filenames anyways.
|
# ? Jan 12, 2022 16:16 |
Dashes for separating major components, underscores for separating minor components.
|
|
# ? Jan 12, 2022 16:45 |
|
self-help_book.rtf
|
# ? Jan 12, 2022 16:58 |
CPColin posted:self-help_book.rtf Someone has named a few AD groups with the opposite pattern (underscore separating major components, dash separating minor components) at work, contrary to everything else in the domain, and it murders my brain every time I see those names. quote:self
|
|
# ? Jan 12, 2022 17:18 |
|
Is that a manual for the Self programming language?
|
# ? Jan 12, 2022 17:25 |
|
nielsm posted:Dashes for separating major components, underscores for separating minor components. this but the other way round
|
# ? Jan 12, 2022 18:45 |
|
nielsm posted:💩 for separating major components, 🍑 for separating minor components. ftfy
|
# ? Jan 12, 2022 18:48 |
|
and you're just gonna trust that everything handles emoji correctly? exceedingly brave
|
# ? Jan 12, 2022 18:55 |
|
Spaces (ASCII 32) for separating words, unit separator (31) for separating minor components, record separator (30) for less minor components, group separator (29) for pretty major ones and file separator (28) for really major ones. It's not like we're using them for anything otherwise. edit: My solution is even 7-bit legal. double
|
# ? Jan 12, 2022 18:59 |
|
redleader posted:and you're just gonna trust that everything handles emoji correctly? exceedingly brave Trust. AHAHAHA. TRUST! OHOHOHOHO! TRUST!
|
# ? Jan 12, 2022 19:11 |
|
ExcessBLarg! posted:You should be using underscores and not dashes in your filenames anyways. How long did you have to train until you managed to be so wrong?
|
# ? Jan 12, 2022 19:22 |
|
1337JiveTurkey posted:Spaces (ASCII 32) for separating words, unit separator (31) for separating minor components, record separator (30) for less minor components, group separator (29) for pretty major ones and file separator (28) for really major ones. It's not like we're using them for anything otherwise.
|
# ? Jan 12, 2022 19:53 |
|
Unix filenames are a dumpsterfire anyway and essentially nothing handles them correctly. They are arbitrary binary blobs, not text. Fun prank: put files on someone's system whose name has bytes that (if interpreted as text) have newlines, spaces, invalid unicode sequences, valid but not normalized unicode codepoints, and some invisible codepoints
|
# ? Jan 12, 2022 21:05 |
|
redleader posted:and you're just gonna trust that everything handles emoji correctly? exceedingly brave No, the point of using emoji everywhere is to break all the things that don't handle it. The rise of emoji has been pretty effective at getting Americans who otherwise didn't give a gently caress about supporting other languages to at least try to support unicode.
|
# ? Jan 12, 2022 23:26 |
|
bolind posted:Now, class, when having a simple utility written in bash, which takes one or more filenames as arguments, why is it a bad idea to have a built-in help that goes like this: Bash has had argparse forever lmao.
|
# ? Jan 13, 2022 01:53 |
|
1337JiveTurkey posted:Spaces (ASCII 32) for separating words, unit separator (31) for separating minor components, record separator (30) for less minor components, group separator (29) for pretty major ones and file separator (28) for really major ones. It's not like we're using them for anything otherwise. The smiley faces for denoting temp files
|
# ? Jan 13, 2022 02:28 |
|
I believe the technical term is 'Urist'.
|
# ? Jan 13, 2022 03:11 |
|
|
# ? May 14, 2024 02:22 |
|
DoomTrainPhD posted:Bash has had argparse forever lmao. Absolutely, and same for Python. A very good question to ask yourself before even typing a single character is: "Is this likely to be a solved problem?".
|
# ? Jan 13, 2022 09:57 |