|
or you can use the ~advanced php cache~ which is free and somehow not part of the language
|
# ? Jun 3, 2013 23:41 |
|
|
# ? Jun 11, 2024 16:26 |
|
people bitch about pyc files?
|
# ? Jun 4, 2013 08:50 |
|
MononcQc posted:So yes in production you'll probably want something that restarts software, reboots servers, or allows you to add or remove them, because otherwise you'll slowly end up with not enough machines to keep going. But you won't fall in the case where, say, the connection between servers falls in a bad state and a 3 hours timeout resolves it later on and all of a sudden one end thinks the other was crashed but the other just thought things were slow. I think Paxos has the requirement that a machine that goes down can be brought back up in the same state it was in before it fell over. Faulty RAID controllers do not seem to be what it's designed to cope with.
|
# ? Jun 4, 2013 11:53 |
|
Ronald Raiden posted:people bitch about
|
# ? Jun 4, 2013 11:54 |
|
my stepdads beer posted:or you can use the ~advanced php cache~ which is free and somehow not part of the language its not part of the language because then zend would have a harder time selling PHP Enterprise Edition or whatever the gently caress its called
|
# ? Jun 4, 2013 14:41 |
|
Ronald Raiden posted:people bitch about pyc files? If you change the name of a file in Flask, delete the previous files pyc because it'll gently caress stuff up.
|
# ? Jun 4, 2013 17:03 |
|
Zombywuf posted:I think Paxos has the requirement that a machine that goes down can be brought back up in the same state it was in before it fell over. Faulty RAID controllers do not seem to be what it's designed to cope with. Paxos has implicit mechanisms to deal with new nodes joining the quorum (you can use Paxos to configure Paxos), where it can play catch-up. This was covered in the Google paper about using it in Chubby: quote:5.1 Handling disk corruption The catch-up mechanism exists because nodes in multi-paxos may lag behind: quote:Each replica maintains a locally persistent log to record all Paxos actions. When a replica crashes and subsequently recovers, it replays the persistent log to reconstruct its state prior to crashing. Replicas also use this log when helping lagging replicas to catch up. The mechanisms for this are not in the google paper, but something as simple as submitting a no-op write for an existing entry (I no-op write entry 1, members of the quorum should answer back with the value that was accepted already) are implicit to the protocol. This is likely slow as you may need to replay the entire log history, but Paxos doesn't directly care about speed. Optimizations would likely have you handing off logs or a snapshot from another member on a side-channel, but I don't recall that stuff in either of the original papers (the most that is mentioned from memory is keeping a quick log of most recent transactions for network blips, not full on corruption).
|
# ? Jun 4, 2013 18:30 |
|
MononcQc posted:Paxos has implicit mechanisms to deal with new nodes joining the quorum (you can use Paxos to configure Paxos), where it can play catch-up. This was covered in the Google paper about using it in Chubby: Well this is the sort of buttressing I was talking about. It seems you need some sort of reliable database underlying Paxos to make it work operationally.
|
# ? Jun 5, 2013 13:53 |
|
Yeah, it assumes reliable storage is there, can deal with failures if it can detect them locally, but does not necessarily need persistent storage. It's not always necessary if you decide to go for the crash-stop model: once the node knows it crashed, it wipes its state clean and recovers from peers. That's what Scalaris does, although I couldn't say how well it fares in production given how few people use it.
|
# ? Jun 5, 2013 14:37 |
@ID_AA_Carmack: The lack of strong typing in lisp is freaking me out, even in tiny programs. Writing an industrial program like this seems... unsound.
|
|
# ? Jun 5, 2013 16:45 |
|
gucci void main posted:@ID_AA_Carmack: The lack of strong typing in lisp is freaking me out, even in tiny programs. Writing an industrial program like this seems... unsound. strong typing is the programming equivalent of big government therefore carmack should be all over loose typing
|
# ? Jun 5, 2013 16:50 |
|
the invisible bit of the loose type
|
# ? Jun 5, 2013 16:50 |
|
youre mother seems to be a loose type
|
# ? Jun 5, 2013 18:27 |
|
gucci void main posted:@ID_AA_Carmack: The lack of strong typing in lisp is freaking me out, even in tiny programs. Writing an industrial program like this seems... unsound. Lisp is strongly typed tho
|
# ? Jun 5, 2013 22:37 |
|
VanillaKid posted:Lisp is strongly typed tho dynamic typing doesn't really count as typing
|
# ? Jun 6, 2013 00:05 |
|
Zombywuf posted:buttressing JewKiller 3000 posted:dynamic typing doesn't really count as typing yes it does. weak typing doesn't count tho
|
# ? Jun 6, 2013 01:46 |
|
scons seems like a nice build system if you're the kind of person who lives in a nightmare where you have to think about build systems. thats my programming story for the day
|
# ? Jun 6, 2013 02:19 |
|
http://www.electric-cloud.com/blog/2010/03/08/how-scalable-is-scons/
|
# ? Jun 6, 2013 02:44 |
|
Deus Rex posted:yes it does. weak typing doesn't count tho enjoy your p-lang hellscape
|
# ? Jun 6, 2013 03:13 |
|
JewKiller 3000 posted:enjoy your p-lang hellscape i didn't say it was goodor that i liked it, just that it counts as typing, which is true
|
# ? Jun 6, 2013 03:18 |
|
Decider('time stamp-match') fixed the worst thing I encountered today, but that was only ~3600 files
|
# ? Jun 6, 2013 03:41 |
|
it's more of a ding on the maturity of scons. old mistakes are new mistakes are old mistakes. you fixed your issue by invoking 'do it like make, goddamn it' the best thing about scons is that you can write things in python. the worst thing about scons is that you have to write things in python.
|
# ? Jun 6, 2013 03:57 |
|
you can still have type errors in a dynamically typed language, its got typing
|
# ? Jun 6, 2013 05:23 |
|
today, i conceptualized a threaded process and it worked as expected on the 1st pass this what god is like being like is
|
# ? Jun 6, 2013 05:38 |
|
but what about the 2nd 3rd 4th and 5th passes
|
# ? Jun 6, 2013 05:39 |
|
Suspicious Dish posted:but what about the 2nd 3rd 4th and 5th passes
|
# ? Jun 6, 2013 05:45 |
|
Suspicious Dish posted:but what about the 2nd 4th 3rd and 5th passes
|
# ? Jun 6, 2013 05:58 |
|
if were talking about things we did that surprised us i was testing out a haskell memoization library today and its like wow i can just write out the recurrence relation i came up with and thats it? felt good
|
# ? Jun 6, 2013 06:09 |
|
MeruFM posted:today, i conceptualized a threaded process and it worked as expected on the 1st pass what about when the device decides to check its email at the same time and the lead horse loses the race?
|
# ? Jun 6, 2013 06:32 |
|
electric cloud makes some pretty impressive build software; they're really good at what they do (they're also really expensive)
|
# ? Jun 6, 2013 11:27 |
|
Suspicious Dish posted:but what about the 3rd 5th 4th and 2nd passes
|
# ? Jun 6, 2013 14:21 |
|
I made that joke first tef
|
# ? Jun 6, 2013 16:50 |
|
Ronald Raiden posted:I made that joke first tef if only we had some synchronization mechanism in place then this tragic joke-after-joke incident would never have occurred.
|
# ? Jun 6, 2013 17:08 |
|
Jonnty posted:enough of the synchronisation jokes shut up, killjoy
|
# ? Jun 6, 2013 17:13 |
|
enough of the synchronisation jokes
|
# ? Jun 6, 2013 17:13 |
|
|
# ? Jun 6, 2013 17:15 |
|
Ronald Raiden posted:I made that joke first tef eventual consistency
|
# ? Jun 6, 2013 17:45 |
|
like ordering is the major prob
|
# ? Jun 6, 2013 17:45 |
|
some group at my work which has a long history of making terrible decisions and insane "we'll just write it all ourselves from scratch" type poo poo just decided to start using scons idgaf what they do or how much time they waste but just for my own amusement what are some better alternatives to scons??????????????
|
# ? Jun 6, 2013 17:48 |
|
|
# ? Jun 11, 2024 16:26 |
|
titaniumone posted:some group at my work which has a long history of making terrible decisions and insane "we'll just write it all ourselves from scratch" type poo poo just decided to start using scons maven? and is scons particularly bad or something? i've never used it
|
# ? Jun 6, 2013 17:51 |