No.534145
1. Test the software, run a testserver yourself
2. Submit errors, bugs
3. Learn javascript kek, javascript is not a serious programming language
4. Implement missing features yourself
No.534156
No.534158
Can we just turn this thread into a Chan engine general?
Between Wakaba, TinyIB (not to be confused with Tinyboard AKA StopNerds.org), and Vichan, whats the best/most stable?
No.534164
No.534167
>>534164
What did happen to the InfinityNext software? Did Josh take it all for himself? Does Jim have it laying on his hard drive being held captive by his pigs? Or was it actually GPL'd and is on Github?
No.534168
No.534171
No.534175
>>534170
I'd rather wait for HWs to adopt Lynxchan than go to Endchan,endchan is full of smug elitists, HWs owns endchans alternative domain though
No.534177
>>534175
Why not just use both? Since 8chan is getting lybx at some point, may as well browse endchan to get used to it.
No.534179
No.534181
the overchan will consume all you're chans
No.534200
>>534158
Actually infinity next is most stable, but not most efficient. Wakaba is most efficient. Strictly of those you listed, none.
No.534202
>>534175
>HWs owns endchans alternative domain
He bought one alternative domain, there's a million more still available.
No.534203
No.534204
What is the problem with the AGPL license for a chan?
No.534206
>>534177
Hotwheels will probably make it look more like current 8chan before he switches over.
No.534209
>>534167
agpl means no soft serve added, no soft serve means no 2ch hosting. 8ch is basically 2ch now as hw will never give up the bang maids and naked gta that 2ch affords him. It's still sitting on the github where it will stay, forever unused.
No.534218
>>534204
Hotwheels said Jim didn't like it, considering that he didn't know what open-source was the problem was probably with Jim. Specifically it was with the ad system and the 8chan (TM) pass needing to be closed source but not sure if they couldn't have been developed in a way that they weren't dependent on Next.
No.534229
>>534203
>16chan
Kek, what is this? Josh just started his own chan site altogether with Next software?
No.534231
>>534200
>Next
>most stable
It sure as fuck wasn't when Josh redirected our /v/ to the Next board during the public beta, even if you consider it having limited resources it was buggy as fuck and was DDOS'ing itself because the database was so poorly designed
No.534234
>>534218
2ch is in the top400 websites, to put into perspective, 4chans rank is 900 and 8chs is 12k, being part of 2ch gives 8chan a competitive edge, especially considering 8ch is linked right there on 2chs BBS page
No.534235
>>534200
Also
>Wakaba
>most efficient
>perl
No.534244
>>534235
Perl has fucking amazing string handling performance, it wouldn't surprise me if it did well.
No.534245
>>534229
It was a domain he and Hotwheels registered after the "shut down 8chan" incident that I guess he decided to use to try and continue Next.
No.534271
>>534244
I thought Perl was slow as fuck as a CGI script compared to php though?
No.534277
>>534167
It's left to bit rot on Josh's github. He hasn't touched it in almost 2 weeks now.
No.534295
>>534175
Well, anything other than /am/ is relatively slow on endchon, but we have *some* quality. C'mon and take a peek, we're not that smug.
No.534333
No.534345
>>534271
Overall probably but for a string heavy website like a chan perl could easily be adequate.
No.534349
>>534271
And this is why nobody competent has used CGI for the past 20 years. With FastCGI that slowness does not exist, PHP can only clutch at straws with shitty caching layers that never work fully.
No.534350
STEPHEN
I'M SORRY WE EVER DOUBTED YOU
PLEASE
WAKE ME UP INSIDE
(I CAN'T WAKE UP)
WAKE ME UP INSIDE
(SAVE ME)
No.534423
>>534137
>node.sjw
hopefully never
No.534553
>>534423
>>534425
sup, josh
No.534734
>>534231
The database isn't the bottleneck with Next, you fucking retard. Go back to /v/ instead of opining on shit you don't understand.
No.534833
No.534836
>>534734
>>534423
>>534425
>asshurt
>sages on every post
>next fanboy
You could learn a thing or two about subtly, josh.
No.534843
is this site being abandoned just because of the josh fuck up? just set up a lynxchan instance for us to mess around with and contribute code to to see if we can get a community driven upgrade to this broken fucking site.
No.534844
https://www.youtube.com/watch?v=QrxPuk0JefA
^ Simulation of Null getting kicked out of development.
No.534872
>>534209
That's just plain fucking wrong.
You could build an abstracted ad serving system that takes into account multiple ad providers. Hotwheels even said that josh was working on an abstracted ad serving system which would allow them to interact with soft serve via oauth.
No.534873
>>534872
Well, you would know, Josh. But that's in the past now you aren't working on anything, and are actively hostile toward 8ch and hotwheels which is the real reason Infinity Next will never be used.
No.534875
>>534836
Hi josh.
>>534843
Lynx is incompetent when it comes to web development. We've had multiple threads making fun of him. I can't wait for it to burn down faster than Next did because cripple can't fix the vichan problems and has pretty much abandoned VFEC (something that'd probably have worse performance than next without document caching.)
No.534879
>>534734
I thought the db WAS the bottle neck? Its' bulk required too much processing power.
No.534882
>>534873
He's probably hostile because hot wheels tried his stupid personal army bullshit on him. Hotwheels doesn't even want to develop or admin this shot hole anymore, wonder why he was pushing for Josh to migrate everything alone, without his help?
Because he was going to give him admin privs, and basically did by giving him access to the db and load balancer.
Something he said he would never do.
But I get it, he's got his health problems. Can't fault him, but you can fault him for his poor choices even though he's probably got chronic pain prescriptions that rip him to his tits.
No.534884
No.534888
>>534879
Templating from db is the bottleneck. Db side is handled well, as expected with using an ORM, except for maybe performance implications of attributes using thr ORM which can be turned off or optimized.
No.534898
>>534882
I agree, we can blame him. He was given fair warning that josh was an unstable faggot that stalked and raped that girl from kiwi farms due to his mental condition. I honestly don't blame josh though, he's just mentally ill and we were wrong to rely on him.
No.534906
>>534898
It wouldn't have been a problem except hot wheels skill as a project manager sucks ass and wasn't focused on integration beyond feature parity. Keep in mind that failures in overseeing a project is also why board stats ballooned to the point that it did as well, though codemonkey (Ron) missed that fuckup as well, even though josh:s pull request clearly stated how the new board stats worked.
No.535503
LYNXCHAN WHEN
The server took too long to submit your post. Your post was probably still submitted. If it wasn't, 8chan might be experiencing issues right now -- please try your post again later. Error information:
No.535504
LYNXCHAN WHEN
The server took too long to submit your post. Your post was probably still submitted. If it wasn't, 8chan might be experiencing issues right now -- please try your post again later. Error information: The server took too long to submit your post. Your post was probably still submitted. If it wasn't, 8chan might be experiencing issues right now -- please try your post again later. Error information:
No.535571
>>534175
>smug elitists
>being this anally ravaged
What a cuck. Low test, low energy cuck.
>forgetting how smug and elitist 8chan is against cuckchan faggots
>>535504
>>535503
>>534137
HOPEFULLY FUCKING NOW. THIRTY FIVE FUCKING TRIES TO POST ON /pol/ WITH A WEBM AND EVERY FUCKING ONE FAILS
Someone start spamming the fuck out of the nigger's IRC and twitter accounts and get his lazy crippled useless ass out of emojinigger land and fix the FUCKING site.
No.536537
>>535571
>THIRTY FIVE FUCKING TRIES TO POST ON /pol/ WITH A WEBM
It is getting annoying as hell trying to post webms to /pol/. So many posting errors.
No.536600
>>536537
maybe you should go outside
No.536614
FIX 8CHAN KRIPPLEKIKE
REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEeeeeeEEEEEEEEEEEEEEEEEEEEEEEEEEEEeeeeEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEeeeee
No.536622
Goddamn HW and Jim
You could at least run a test board with lynx.
No.536691
>>536622
I'd be interested to see how it performs under the same circumstances given to Josh. Create a beta domain, migrate data, redirect boards at it, see what happens. Better than doing absolutely nothing because
>lel /pol/ 1.5k PPH everything is fine guise
While we get 10 posting errors per webm.
No.536800
>>536691
It'll perform just fine when streaming to socket because its just serving static files. Exceptionally, most likely. Problem is potential rebuilding problems. How fast node can serve from gridfs isn't really an indication of how the site will perform. Because threads are served the same way with this website, and has similar performance characteristics (if not better with nginx, assuming they're in kernel cache) and yet we have rebuilding and database problems.
Let alone the fact that it shouldn't be serving documents like threads or json or RSS from gridfs abstraction but that's another fuck up.
I think that he should set up a test server because it'd be hilarious to see everyone sperg out about how shitty 8tailedllynx is, way worse than next seeing as those are the only things even /tech/ can give opinions on considering how fucking stupid/incompetent people here are.
Oh well, its the grave the cripple chose.
No.536803
>>536622
Hotwheels doesn't want to admin this shithole anymore. He's fucked up so many times, not even considering the embarrassing situation where he fucked up as a project manager with josh because of his ego, that he's afraid to try. And because of his ego he halfheartedly endorses some piece of shit that made pretty much makes the same mistakes that tinyboard/vichan made, and more that it did not make (lynxchan mod panel vulnerabilities for example.)
But its totally okay, because these architectural mistakes were made by someone working for free. So sayeth the cripple because he wants to piss off some mentally unstable faggot and try to preserve his ego.
No.536868
What are potential considerations for load testing? What test cases would need to be included to accurately reflect site usage on a site like 8chan?
I have these "test cases", but obviously there are more:
* Display home page
* Display board list
* Display board frontpage
* Display thread
* Reply to thread
* Reply to thread with image
* Reply to thread with webm or other video
* Display board page > 1
Can anyone help compile a list?
Or can someone key in to the most commonly used pages are?
If applications like lynx chan can be load tested, there is a possibility that they will survive. At least it will validate that whatever Yet Another Image Board fucking works worth a shit.
No.536870
This might be retarded but why can't we just use whatever halfchan is using?
Moot is hiding the code in her Gf's son locket or something?
No.536900
>>534175
> Endchan,endchan is full of smug elitists,
Oh, and 8chan isn't?
No.536902
>>534137
How many months do we have to wait before Hotwheels fixes this shit?
No.536903
>>536870
It's closed source.
No.536904
>>536868
make new thread, delete reply, delete thread, delete file
No.536917
No.536919
>>536622
You do relies there working on that right Lynx chan is not finished when it's done /cow/ a the test for the software then /b/ ect
No.536924
>>536919
English not your first language?
No.537002
No.537005
>>536868
You won't be able to easily load test lynxchan in a way that'll lead to problems.
But endchan was already having problems with slow rebuild times. Its like saying that you're benchmarking tinyboard, and tinyboard is the fastest image board and won't have any problems with load, when serving thousands of requests a second easily when in reality you're just benchmarking nginx serving static HTML files.
No.537010
>>537005
>But endchan was already having problems with slow rebuild times.
>implying
No.537011
>>537010
It was, read the irc logs. Hotwheels himself was complaining, leading into a discussion of how badly lynx had fucked up by repeating the same mistakes vichan made and also lynx lying about gridfs and atomicity, saying that (mongodb) gridfs prevents the locking issues we've seen, when it does not when you are using gridfs at least.
No.537015
>>537011
I read the logs, couldn`t find anything there besides odili abusing rebuilds and expecting nothing wonky to happen.
No.537017
>>537015
Wonky in this case being slow, slow global rebuilds.
No.537021
>>537017
>I need to get a beer.
>I know, I'll drag the whole fucking fridge to the living room to get a beer
>Man, this fridge sure is heavy, if it were lighter, it would be much easier to get my beer
Yeah, w/e floats your goat.
No.537030
>>537021
Global rebuilds are things that need to happen regularly in response to front end changes and global announcements with your architecture.
Its okay lynx, at least people aren't making fun of you for CSP or CSRF in IRC. Wouldn't want to damage your fragile ego, you being the type who has all the answers.
No.537056
>>537030
>Global rebuilds are things that need to happen regularly in response to front end changes and global announcements with your architecture.
>Engine allows to rebuild exactly what you need instead of the whole damn thing
https://gitgud.io/LynxChan/LynxChan/blob/master/src/be/Readme.md#application-usage
But fuck reading, right?
No.537067
>>537056
And you probably think your rebuild queue isn't going to turn out to be a bottle neck as well.
No.537073
>>537056
How does that change anything? You either need to fix global and board announcments and triggers to use your queue and only modify parts of the DOM that it needs to modify. Its a problem in both cases. In response to front end changes itd still need global rebuild. And i doubt that your queue, even only modifying the part of the document that needs modified, wont still shit the bed when you throw thousands of boards and all of the threads on then at it when making a global announcement.
No.537100
>>537067
It's web scale because opinions, reading and node js with mongodb. Obviously.
No.537102
>>537067
nntpchan had this exact problem before. When we got hit with the 8ch.net firehose during that end of 8chan event that redirected to topchan and i2p.rocks. the build queue lagged to 90 seconds with the sheer volume of regeneration requests.
The Problem should be lessened by a lot but it's still going to be a big issue at 8chan tier post volume.
No.537118
>>537102
The solution would actually be quite simple (at least in your case). Instead of marking pages for rebuild on events you just invalidate the cache, because it seems to me that your cache is already capable of handling cache misses anyway:
https://github.com/majestrate/srndv2/blob/master/src/srnd/redis_cache.go#L187
I don't think this could be done as easily in lynxchan.
No.537130
>>537118
there's a mechanism already in place in nntpchan's code, a caching abstraction.
not sure how stephen's node-pile will handle that besides praying to the memegod
No.537137
>>537067
It won`t because the rebuild queue drops unnecessary rebuild pages.
Not only that, but when you run lynxchan as a cluster, the slaves handle the rebuilds instead of the master, so once your chan grows enough to require multiple servers, each core of each slave will be able to process messages.
So no, this issue was solved a while ago.
No.537140
>>537118
I might as well use redis for pages instead of mongo if I am going to do anything different, since doing what you propose involves making several threads across several servers to handle this cache.
What happens in your scenario when two requests are made for the same page that doesn't have a cache, is that both will trigger a rebuild for it.
And that would be a huge DOS attack vector.
Just request pages like crazy and keep posting on them to invalidate the cache.
No.537144
>>537140
> What happens in your scenario when two requests are made for the same page that doesn't have a cache, is that both will trigger a rebuild for it.
If there's a cache miss while rebuilding it rebuilds twice, duplication of work isn't that horrible given this edge case.
> And that would be a huge DOS attack vector.
> Just request pages like crazy and keep posting on them to invalidate the cache.
Sounds like a normal day's activity on 8chan
No.537184
>>537140
Then your software is already fucking naive if it doesn't have flood limits/antispam. How the fuck you've convinced hotwheels that your software is worth two shits is anyone's guess, its clear that there are very fundamental issues that require going back to the drawing board. In a way that's not just a major version increment.
No.537200
>>537137
How does your rebuild queue 'drop' unnecessary page rebuilds when all threads and indexes need rebuilt in response to a global announcement, or board wide rebuilds for board specific messages. You're still going to have a massive rebuild queue.
>muh sharding
Right
No.537211
>>537184
>Then your software is already fucking naive if it doesn't have flood limits/antispam.
Which it had since last year. Which I included more features for that and improved the existing ones continuously.
Look, you are either malicious or irreparably ignorant.
Have fun sperging about shit that anyone would be able to see its non-sense within 5 seconds of using the software or reading the documentation, I am not wasting my time with you until you actually learn what you are talking about.
Currently you are no different than a chess match against a pigeon.
No.537221
>>537211
Then there's no DoS risk, you fucking retarded monkey. Invalidation should only happen in response to successful posts, preferably fired in response to event/listeners.
Its more like are arguing with some retarded monkey who can't explain how his rebuild queue won't shit the bed when suddenly flooded with hundreds of thousands of threads, and tens of thousands of index pages that need rebuilt. You keep repeating the same shit over and over again, that your rebuild queue or mongo takes care of everything, when it does not.
No.537224
>>537221
Why haven't you touched infinity next in 15 days Josh?
No.537228
>>537224
Nice meme friend, but I'm not josh. You're free to explain how lynx isn't a fucking retard though.
No.537231
>>537228
>my claims are true until proven otherwise
top sperg
No.537233
>>537231
You confirmed that global rebuild perf was shit on endchan or on whatever server, response to global rebuild "abuse."
Which shouldn't be thought of as abuse, because rebuilding of indexes and threads themselves needs to be done with your architecture in response to global announcements, front-end changes and potentially xss fixes, in the latter needs to be done as fast as possible.
With any other document cache, you would simply invalidate and recache in response when needed.
No.537235
>>537102
>nntpchan
nice vaporware
No.537236
>>537233
me, who?
I just came here to call you a sperg.
No.537237
ITT we are angry about things we don't understand
No.537238
No.537239
>>537238
I never pretended to be making one, you sperg.
No.537240
>>537239
Then don't fucking reply to a conversation that was between another anon, myself and some retarded monkey who thinks discarding rebuild triggers makes his rebuild queue magically able to handle hundreds of thousands of rebuilds quickly when it can't even handle rebuilding endchan quickly
No.537244
No.537248
>>537244
If you're going to use php you might as well contribute to next despite hotwheels ego. Features blow that out of the water and no framework means more code needs tonne written to abstract potential choices admins can make. Also doesn't even have document level caching support or board creation, as far as I can see, let alone sane user management or board owner options.
No.537254
>>537248
Also rebuild queue stored in database instead of using events and such. Sessions use simple tokens, no protection against xsrf. Localization will be worse than vichan. Worse than lynx tier.
No.537261
>>537254
>>537248
Still a huge WIP, ages away from parity with infinity. I believe we have a slight edge on lynx, feature wise. Performance has been relatively untested due to nobody knowing we exist. It's currently tooled towards traditional chan setups, one board at at time. We're slowly retooling it to support user boards.
Might take out an ad and ask anons for help load testing later down the road.
>>537254
what do you mean by document level caching?
I do appreciate the feedback anon(s).
No.537262
>>537261
Caching of whole pages via passthru with varnish or something like redis instead of a complete retemplate every request.
No.537263
>>537262
>>537261
But actually it doesn't apply because presumably you're caching on disk already with your rebuild queue. I was looking at it and posting both on my phone, and thought it was completely dynamic.
One thing which you could use to differentiate yourself is move away from Unix timestamp filenames and instead use sha512 lookups. Hashing sha512 will be faster than sha256, is better than relying on md5 in which collisions can be created in 30 mins on a desktop.
No.537430
>>537235
it works and has worked for 2 years now. no one uses the infrastructure but it's definitely not vaporware.
No.537454
>>537240
What if I refuse? You are going to spergout even HARDER?
:^)
No.537456
>>537261
>nearly 1k commits
>huge WIP
Excuse me?
No.537458
>>537263
I agree on sha512, it allows for easier file dedup and does not have the realistic potential for duplicates like with timestamps (unless you use nanosecond granularity huehuehue)
No.537534
>>537263
>>537458
I'll look into tweaking it then, thanks!
>>537456
We're designing it to be drag+drop, edit 4 config lines and run, all while offering everything something like infinity has to offer. We're still a while away from that, which is why I say WIP
No.537625
>try to report cp
>hurr durr 404
>maybe already deleted
>go back and refresh
>still there
>try to report again actually get report form this time
>report
>lel error 503
JUST
No.537636
>>537625
MEMEGOD DATABASE
No.537656
>>537636
He was talking about 8ch, though.
But nice meme.
No.537702
>>537656
This is correct, I did mean 8ch.
The server took too long to submit your post. Your post was probably still submitted. If it wasn't, 8chan might be experiencing issues right now -- please try your post again later. Error information:
No.537727
No.537731
>>537727
>slower than PHP
>fixing anything
Sure thing, m8.
No.537734
>>537731
>slower
wew lad did you even read the op or do you mindlessly follow what other people say
No.537735
No.537736
>>537727
It's interesting, but he shouldn't have used python. Maybe Golang instead. I hope he has great success, and he pitches his site to Hotwheels and it BTFOs Infinity Never in the benchmarks enough that we can use it. Something needs to be done we can all agree on that. I don't think python is the answer to our problems though.
No.537738
>>537735
Hey man, the performance of the programming language does not directly relate to the speed of the website. I think you should try to post on the demo site and read on /test/ how it works.
>>537736
Again, the raw performance (in benchmarks even) is just not that important. What's important is a good infrastructure and logical caching.
The real performance benchmarks are thing like posting speed and read speeds, and like demo-ed on /test/ these are pretty good.
https://uchan.plebco.de/test/view/271
No.537740
>>537738
Except that DOM manipulation is extremely CPU intensive.
So in this case, the performance of the runtime environment matters a lot.
No.537743
>>537740
>DOM manipulation
Except that there is no dom manipulation?
Do you mean template rendering? Because that's no real issue, especially with varnish. Please see for yourself on /test/
No.537749
>>537743
Of course that it won`t be noticeable on such a small scale.
Each page could take 500ms to be built and it wouldn't raise any suspicion.
Now put that on any considerable load and your gears will stop.
Remember next? Where the beta would run mostly fine and then when it had to handle the actual load it would crawl to a halt?
No.537810
>>537749
Next had no document caching whatsoever in beta. It had an internal httpcache on the index as a test. That's as far as it went.
No.537970
No.538030
>>537240
>monkey
Josh confirmed
No.538033
>>538030
>Implying Brazilians are human
No.538034
No.543443
>>537749
>>537743
I thought lynxchan created static documents for nginx to serve? Since the page rewrites are only done on posts and reads don't even touch node isn't saying DOM manipulation is CPU intensive kind of a misnomer?
No.543457
>>543443
Reads do touch node, it has to stream to socket.
But I'm pretty sure Lynx's piece of shit creates a document from a DOM representation and writes it to file, or so he implies. It doesn't matter, how he's storing the file is not concurrency safe.
No.543572
>>543443
It doesn't use webservers.
The engine creates HTML pages, stores them on mongo and the engine itself serves them.
No.543597
>>543433
There was a bug that would crash the engine if the data for location flag wouldn't be there.
They have installed the location data and I have fixed the bug.
No.543642
After HW stops programming his EEMOGEES
No.546805
but mongod doesn't drop things D:
No.546831
>>546805
What is this even supposed to be showing?
No.546928
>>546831
It shows that using memgod for storing literally fucking everything is a terrible idea because shit will randomly disappear because memegod isn't reliable.
No.546947
>>546805
>posts missing at 5PPH chan
wew
No.546967
No.546973
>>546928
>not wanting 8ch to be webscale
anon pls
No.546976
>>546947
W E B S C A L E
E
B
S
C
A
L
E
No.547092
>>546976
plz stop bullying lynx, everyone knows that he's competent after hotwheels mentioned that lynx did everything better than josh for free in his passive agressive shitpost of an article.
lynx is web scaling as best as he knows how, plz no bully.
No.547125
>>547092
lynxchan dev pls go