DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2021-10-01 00:00:00Z and 2021-10-02 00:00:00Z

xD

@timakro what about russian unique servers? im glad to host them too

With russian discord unique race server too? poggers

poggers

nah

sad

06:42

No unban for me feelsbadman

feelsbadman

[ddnet-web:master] 1 new commit

edd9e51 Frontpage video - def-

fokkonaut

@Learath2 how is secured.gg doing with USA and GER?

Not so good yet.

08:19

Trying to work with deen on improving the handshaking a little, that would make filtering a lot easier.

08:19

However I'm onto something atm that should work pretty good

yay breaking protocol changes

No.

:)

Handshaking itself wouldn't be changed, I suggested introducing a "force reconnect packet"

08:20

however, I can also just abuse the "server is full" reconnect, which of course isn't transparent, but would stop pretty much anything currently leaking. (edited)

Im not sure what you are on to tbh :D

-> validate handshake using a custom challenge

08:21

-> whitelist flow, reconnect player and allow traffic to acutally pass through to the game server

How much incoming traffic is guaranteed on secured.gg?

Attack ,or clean trafifc?

both

Clean traffic just as much as the actual server can handle of course.

08:22

which in this case, is 10Gbps in both USA & DE

08:23

Attack traffic? Depends. Amplification attacks generally don't cause any trouble, and are filtered upstream. L7 (game) filtering 100Gbps, can however be upgraded pretty fast on our end. (edited)

08:23

We're aware

08:23

that's no problem

08:23

It's the first handshake packet which is being sent, which is not challengeable or cachable on our end. (edited)

08:24

which makes things a little problematic.

08:24

However, introducing a reconnect packet to implement a custom challenge, should work perfectly.

08:24

which would be seamless for the client

08:25

"first handshake packet" being "\x10\x00\x00\x01TKEN\xff\xff\xff\xff"

yea but game traffic doesnt use tokens iirc

08:25

only in 0.7

Ok, that doesn't really matter

08:26

We don't expect game traffic before handshaking

true

08:26

but if he sends spoofed packets with ips from connected players

A flow consists of

08:26

iph id, source ip, source port, dest port, dest ip

yea

Chances are a few packets will then still leak due to them perfectly matching that

08:26

but it'd be so low volume, it can't cause issues.

08:28

and you can of course also rate limit that traffic, even though I'm not a huge fan of rate limits

problem on my server right now is, that OVH limits incoming traffic to a very low amount during an attack, that means the custom firewall cant filter all packets because also legit packets are being filtered

08:28

I'll move soon, then its not a problem anymore

08:28

great to see other servers working on a solution too

If @deen can implement what I suggested, it shouldn't be a huge issue for servers that are with us

08:29

If he can't implement that, I'll abuse the reconnect mechanism that already exists, also not a huge issue. (edited)

Just use that I think

08:29

older clients are still around

08:29

anyways^

It'll display that "server is full, reconnecting in 4 seconds" message

08:30

so it wouldn't be super transparent

08:30

but it'd work

It works

09:07

09:08

<idle>-0 [004] ..s. 19300946.505675: 0: generated and saved challenge: 454029306 <idle>-0 [004] ..s. 19300946.644496: 0: success! <idle>-0 [004] ..s. 19300946.644505: 0: yes

@Fän do you mind elaborating on what is missing on ddnet's initial token handshake? Why do you need a custom challenge?

denkbar

@Fän do you mind elaborating on what is missing on ddnet's initial token handshake? Why do you need a custom challenge?

Nothing's missing, to be fair. (edited)

09:43

Ideally, our filtering environment should be able to issue challenges, which it is. However, as of right now, we can't make the server "state-aware".

09:43

For this we'd a) require new gear, which is coming in some time october or b) simply make the player reconnect.

that's interesting. Well, I hope it works

Try connecting to a GER1 server and you should notice how it'll reconnect you (edited)

yea, it does

Only happens once though.

so, at that point my ip address is whitelisted?

The flow, not the IP directly

09:47

there's more to it, but yes, your client is whitelisted at that point.

the thing is, when I open a second client, the dialog also doesn't appear again

Agreed

@Fän how does that thing work? I currently have a website https://f.zillyhuhn.com running to validate an ip address via tcp handshake, could that be replaced by ur reconnect thing?

If you're asking if you can run this with your dedicated server / vps, then likely not

wym?

It uses xdp / eBPF

whats that

https://en.wikipedia.org/wiki/Express_Data_Path

Express Data Path

XDP (eXpress Data Path) is an eBPF based high performance data path merged in the Linux kernel since version 4.8.The idea behind XDP is to add an early hook in the RX path of the kernel, and let a user supplied eBPF program decide the fate of the packet. The hook is placed in the NIC driver just after the interrupt processing, and before any mem...

and why shouldnt i be able to use this?

VPS gets killed even with XDP xd

Not all NICs support this, especially not in a virtualized environment.

Not on my vps

And even then, there's the capacity issue.

but the firewall infront

If you have a 1Gbps link, and that gets blasted, there's nothing you can do

I

10:29

k*

fokkonaut

but the firewall infront

An interesting article about that from CF https://blog.cloudflare.com/how-to-drop-10-million-packets/

How to drop 10 million packets per second

Internally our DDoS mitigation team is sometimes called "the packet droppers". When other teams build exciting products to do smart things with the traffic that passed through our network, we take joy in discovering novel ways of discarding it.

We do all our filtering in XDP / eBPF, and it's great, however not everyone can just "set that up"

I know, but I know someone who could

10:30

can you explain me how it works, maybe in dm

Fän

Ideally, our filtering environment should be able to issue challenges, which it is. However, as of right now, we can't make the server "state-aware".

^

idk wym

10:31

Or in what relation this is

I've explained how the whole thing works a few msgs above

ah

however, open sourcing this as example won't work

10:32

as your server still can't handle > 1Gbps or whatever its capacity is, and even if we did open source this or smiliar, attackers will be able to find a way around.

10:32

Good luck!

@Fän maybe instead of disconnecting with "full" you could instead display something like "Please wait..."

denkbar

@Fän maybe instead of disconnecting with "full" you could instead display something like "Please wait..."

not possible

^Yes possible

Not how it is rn

deen already mentioned it, and I've made the changes

ah, okay

just need to push it live.

Isnt that message clientside?

Not at all.

Oh didnt know

That's how we force the reconnect

ah, the client uses the keyword "full" for detecting whether it is full

10:38

now, everything makes sense

denkbar

ah, the client uses the keyword "full" for detecting whether it is full

yea

denkbar

ah, the client uses the keyword "full" for detecting whether it is full

correct.

10:40

            if(g_Config.m_ClReconnectFull > 0 && (str_find_nocase(ErrorString(), "full") || str_find_nocase(ErrorString(), "reserved")))
                m_ReconnectTime = time_get() + time_freq() * g_Config.m_ClReconnectFull;

hum, "please wait four full seconds" contains "full"

monkalaugh

1

Oh god, hpw did we leave that in there? String matching EW

11:26

s/hpw/how/

Works great for me

I mean it worked out, but still it’s very meh software engineering

Agreed

11:27

but it does help a lot now

So what is the idea btw? How does making people reconnect help?

I explained above

11:28

upon first connection, our xdp mitigation will challenge the connection, instead of the game server doing so (edited)

11:28

once that's verified, it'll reconnect the player to "sync" states

Ah, that’s a nice idea. I had a PoC that used an xdp filter on the server and I instead used a bpf map

Well, bpf maps aren't the issue here

11:31

As we're intercepting the initial handshake challenge, the actual game server won't receive it

11:31

that's why we're reconnecting the player

11:31

to whitelist the flow, we're of course still storing data in bpf maps

Well, I was talking about what I did. I did let he game server receive the initial handshake. I can see how handling that entire flow in xdp would be more efficient

As a generic solution I would be really happy if some hosters provided a way to upload custom xdp into some upstream firewall.

That won't ever be a thing.

11:34

It's a liability that no hoster will willingfully take.

Yeah, too flexible

You can cause quite a lot of trouble otherwise.

Then some reasonable subset language

11:35

I just wish I could configure the firewall without having to ask a hoster to do it for thousands of dollars

Well, there's also the issue of most hosting providers not even knowing what eBPF / XDP is.

11:36

Which is the reason as to why custom ddos protection providers are so successful nowadays.

Fän

Well, there's also the issue of most hosting providers not even knowing what eBPF / XDP is.

Its rather "new"

"new" technology from 2016

Some companies are using JAVA 6

I’m not sure a reasonable subset exists, an upstream firewall would have access to traffic not even destined for you. I could see it maybe being a thing on an “upstream” firewall that is right before your server

Learath2

I’m not sure a reasonable subset exists, an upstream firewall would have access to traffic not even destined for you. I could see it maybe being a thing on an “upstream” firewall that is right before your server

Which no sane provider would allocate >10Gbps to

Yeah, I don’t see it happening. Maybe a very crippled verifiable language that gets compiled down to bpf

Maybe just use "standardized" stateful protocols instead

As in?

11:41

The solution that would work on most hosters without much friction would be to mimic the protocol of some popular game

Even then, providers like Hetzner have basically no protection.

11:43

The only provider I could think of would be OVH, which most servers are still being attacked on (edited)

Hetzner offers ddos protections, but mostly you end up nullrouted

I don’t think hetzner null routed us even once the last 3 years

We got nullrouted twice in 6 months already, not KoG

Fän

Even then, providers like Hetzner have basically no protection.

https://www.hetzner.com/unternehmen/ddos-schutz

11:45

But you right, hetzner isnt the best player out there

Avolicious

But you right, hetzner isnt the best player out there

This is just for amp-related floods

11:45

Anything actually spoofed won't be stopped at all.

@Fän btw did you consider the 2 other protocols we support when you wrote your rules?

"DNS reflection, NTP reflection or UDP floods on port 80."

11:47

basically things that were blocked using static ACLs in 2010 already

Learath2

@Fän btw did you consider the 2 other protocols we support when you wrote your rules?

I haven't tested older clients

11:47

If you're talking about the info request (i don't have a sample of that atm), then yes, that's passing just fine.

Vanilla 0.6 doesn’t have a token, 0.7 doesn’t use the same layout for the header (e.g. the token isn’t in the same place)

11:49

If you are only deploying this filter while under attack it’s fine to drop these, but if it’ll be always deployed we might have to allow those flows through albeit with a ratelimit

Of course I'm not very deep into all this, but isn't it granted to assume that most players would use the newest version?

11:49

(found at https://ddnet.tw)

Yep, a very large percentage of players would be using either the ddnet client or derivative clients that will use our version of the handshake

11:52

But the servers support all 3 versions of the handshake, so we probably should allow those through aswell

11:53

Or if this is only deployed during attacks maybe we should just not advertise on the 0.7 master during that time

I can likely add support for 0.7

11:55

not entirely sure about vanilla 0.6 though

You can just let that through with a ratelimit and let it drop in userland?

Just to get complaints about "server getting ddos"?

11:55

Rather not

11:55

I'll see what I can do

Learath2

You can just let that through with a ratelimit and let it drop in userland?

Let's say a newer client uses the protocol 0.6 uses, however sends their version as newest (afaik a certain packet contains the ddnet client version), will it allow that to pass through?

The server doesn’t really pass things through

11:57

Not sure what you mean

ok, let me rephrase

11:57

Right now, the newest version connects using an initial challenge

11:57

if the client did not send the challenge, but connected smiliar to how 0.6 connects, would it still allow the player to join?

It won’t take up a slot. The server has a “handshake” built for 0.6 clients that abuses a behaviour of the client. So it will get a tad deeper into the netcode then will get dropped

11:59

I can tell you the path it takes in code if you want to take a look

Sure!

So SECURITY_TOKEN_MAGIC iow TKEN is what we use to indicate a client supports the ddnet protocol. In CNetServer::Recv IsDDNetControlMsg will check if the magic is in the packet. If not it will take the 0.6 path. Which is in CNetServer::OnPreConnMsg, this will simulate an accept, there is a nice comment there you can read about what it does

12:07

It basically abuses the fact that the client will return the gametick in it's input packets. So we send a generated gametick, and in the input packet we check that it is the same generated gametick we sent

Cool, can add support for that.

How are you thinking of doing it? Will you re-implement it in XDP or will you pass it through?

Assuming that it doesn't support the same reconnect mechanism

12:10

I'll have to wait for the new gear to come in

12:11

and I then can mess with it, and send custom generated responses to the server and such

12:11

which will make the entire "reconnect" thing useless as we can handle it in an easier way

well, instead of auto reconnect you can also instruct players to reconnect

^^

fan is the guy from secured.gg right?

12:13

i feel grateful but i cant stop thinking about the day he will abandon us

12:13

xd

12:13

if this works out

+1

Hm, what easier way? Don't think storing the handshake packets is viable

Learath2

Hm, what easier way? Don't think storing the handshake packets is viable

It's not.

12:14

As of right now, we reconnect the player to send another handshake packet

12:14

to forward to the server, after validating the actual connection as legitimate

12:14

with the new gear, we can instead generate packets and send them as a "replacement" handshake packet

12:14

that way we won't have to reconnect players at all

Hm, so the firewall/the system in between will do a handshake with the client, if it's a legit connection it'll do another one with the server then let them speak with eachother? I guess that works since we don't have tokens in actual game packets iirc

Correct.

Ryozuki

fan is the guy from secured.gg right?

it's nice to finally have someone that knows what they are doing with xdp and has access to some decent hardware

ye but that doesnt remove my worries xd

That's how our Rust filter works aswell, to some extent

12:16

Obviously there's more to it, but that's one of its main functionalities

@Learath2 do u have access to said xdp code?

We've all been messing around with this kind of solution but with our peasant 1g links it's just too easy to overwhelm

this is rly great, but with such a long lived game as ddnet people come and go you know it

@Ryozuki nope, not sure what the conditions are around that one

its better to be prepared

I'm guessing they keep the firewall stuff a bit of a secret, which is fair, given the nature of this kind of thing

12:23

btw if this solution proves useful we should probably implement a reconnect message so people don't get the "full"

It won't be required with the new hardware anyway

12:24

We're having trouble with a vendor, so things might take another week or two, but at that point we'll have a lot more control over traffic flow

Will you officially support teeworlds/ddnet filters for all customers?

So this whole "reconnect" thing is just temporary

denkbar

Will you officially support teeworlds/ddnet filters for all customers?

Once we consider the filter as fully functional, yes.

cool, I want one

(which means support for 0.6, 0.7 and current version)

but will u need a dedicated or just a vps?

VPS will do just fine

thats nice

We'll probably sooner or later move to 0.7 protocol ourselves

1

So another protocol change?

12:26

anyway, can add support for 0.7 if someone can provide me a download

https://github.com/teeworlds/teeworlds

GitHub - teeworlds/teeworlds: A retro multiplayer shooter

A retro multiplayer shooter. Contribute to teeworlds/teeworlds development by creating an account on GitHub.

Ah I see

ddnet is a mod of teeworlds

12:27

it ended up being more popular

I know, just not fully aware of all versions

ah ok

12:27

well teeworlds changed protocol from 0.6 to 0.7

Well I wouldn't hold my breath, I've been intending to port ddnet to 0.7 for 2 years now. I just don't have the time for it

but ddnet stayed at 0.6

12:27

i wouldnt rush the 0.7 stuff

12:28

this will take lot of time

12:28

and lot of angry people

12:28

monkalaugh

We do support the 0.7 protocol though, the filter should probably allow it

12:28

We just don't use it in the client

porting to 0.7 fully would mean new skin system etc lot of changes and lot of ppl who dislike changes

12:28

xd

@Ryozuki yeah the skin thing is going to be a huge issue

12:29

We'd probably have an extra netmsg to allow people to use legacy skins

also we should never ever take their text rendering stuff

12:29

it sucks ass

12:29

it looks rly bad

Ryozuki

and lot of angry people

ᵃᶜᶜᵒᵘⁿᵗˢ

@TsFreddie @Ryozuki is insulting your text renderer

Image attachment

12:30

cmon

12:30

u know it

12:30

this is rly bad

what's this

teeworlds 0.7 chat

ough

its like it went to 2000's

Image attachment

the skin system is nice tho

12:31

the only BIG downside

12:31

is we got lot of skins

12:31

that would be lost

hm, ours does look prettier, why is that?

@Learath2 ours looks like 0.6 teeworlds

12:31

0.7 changed it

12:31

to use the shadow weird

12:32

and it looks rly bad

nah, not my fault that i kept the look

12:32

but screw Ryozuki anyway

outline > shadow

heh, I found a bug with cl_chat_old funny how I predicted this the second we were forced to include yet another config to keep old players happy

we can remove it

12:33

im sure at this point only konsti and ravie use it

12:33

most ppl are 2 lazy to change

12:33

it

if you turn it on and off you'll get messages without a tee on the left

proximity voice chat ddnet when

@murpi i wanted to look into that

murpi

proximity voice chat ddnet when

that would be so toxic

always wanted to learn voip

12:34

i can see sp someone already

12:34

with a soundboard

12:34

and the nazi anthem

Do I have to modify anything to allow 0.7 players to join?

imagine having to moderate voice chat :/

my server doesn't seem joinable on 0.7

ur ddnet server?

actually, nevermind! (edited)

put map in maps07 folder

Learath2

imagine having to moderate voice chat :/

integrate intels bleep system

@Fän yes, you need to generate the 0.7 map, there is a tool included in the zip for it, and you need to enable sv_sixup

12:35

if there are any issues enabling the compatibility layer there will be a warning on the command line

Image attachment

Fän

iph id, source ip, source port, dest port, dest ip

Q: Restarting the client should be changing the src port, do you think maybe you have a bug there? I only got the double connect the first time

A: Had to limit flows to ip saddr temporarily

TsFreddie

Click to see attachment 🖼️

is this some fake nintendo

why does it have to be nintendo

12:38

its just a arm device with controller

xd

12:39

does anyone want to donate a steam deck to me so i can test ddnet there? (for dev purposes obviously) monkalaugh

monkalaugh

sure y not

our company's publisher got the dev kit

poggers

1

might as well donate to ddnet

don't think there is much that needs to be done specially for the steam deck

12:40

if it works on linux it should work, no?

@Learath2 shh

other than making a better controller support

thonk

feelsbadman

besides why would you donate it to ryozuki, you should donate it to @Deleted User if anything is broken it'll be the gfx

pepeH

i always thought L and R to move and left stick gun right stick hook might be better than using stick to move btw

with a controller its hard to hook + hammer + aim at same time

12:43

something quite frequent in ddnet

Ryozuki

with a controller its hard to hook + hammer + aim at same time

and move

auto aim for switch to make it easier kekw

kekw

well but move would be on the other hand

our ddnet rip off has aim assist lol

lol

12:44

bitcoin got back to 40.000 there goes my hopes for normal gpu prices

imagine using bitcoin

you just tap other's avatar to auto hook. and hook aim will change angle to snap to a hookable surface if it is right at the edge of an unhookable surface.

I feel like switch players will get banned on noby fng guaranteed

hm, following people is still quite hard with the new serverlist, I wonder if we can do anything about that

@Learath2 shareable server links

12:55

the server browser can have a extra button

12:56

"share server link in chat"

12:56

joining should require confirmation

12:56

and it should be heavily rate limited

Hm, how would you click the link tho?

ddnet would provide the functionality

12:56

a special packet

12:56

not an actual link

12:57

just something u can click

That part is easy, I was more thinking UX

maybe add a new tab

12:57

on the esc menu

We dont have a cursor while playing to click with

to see shared servers

12:57

by people

12:57

so its not intrusive in the chat

12:57

and can be easily ignored

12:57

by non interested people

If we had accounts we could do parties. Everyone follows the party leader

eh

12:58

we could do that already

12:58

parties dont need accounts

12:58

well it would only last a single server

12:59

at this point it would be easier to just join a team

12:59

and make a command

12:59

/flw <ip>

12:59

/accept

12:59

gg

Minecraft flashbacks.

i find it amusing you add a dot to all sentences

The server dictating the next ip to contact is dangerous

but its just me maybe

13:00

xdd

Ryozuki

i find it amusing you add a dot to all sentences

Habit

Ryozuki

/flw <ip>

Actually this could work

13:03

flw would send a netmsg to all clients in the team, accept would be a clientside command

13:03

accept wouldnt even be a command actually, just a popup

@Fän lol searching about people who add full stops on instant messaging apps i found this https://www.theguardian.com/technology/shortcuts/2015/dec/09/science-has-spoken-ending-a-text-with-a-full-stop-makes-you-a-monster

Science has spoken: ending a text with a full stop makes you a monster

A new study suggests that people who finish messages with full stops are perceived as insincere. Let’s not get started on the creeps who put a nose in smiley faces

13:03

the internet is crazy

Oh no

“send nudes.”

Linguist reveals why the punctuation mark is making you look 'passive aggressive'

13:04

ye thats what i felt

13:05

about ppl who add dots in msges

13:05

xd

13:05

deen does it sometimes too monkalaugh

monkalaugh

Just looks better in certain sentences, no? (edited)

not for me xD

Surely does for me

i specially get nervous when ppl say

13:06

No.

13:06

xd

Your opinion is invalid Ryozuki.

monkaS

1

idk if they angry or smth

13:07

@Learath2 i know ur on mobile cuz u capitalize ur text

13:07

leeks

You can actually tell exactly which device I'm on at all times

13:08

My ipad will turn :P into

2

Tell me which device I'm on?

ur probs on mobile

13:08

but idk

We havent known you long enough yet @Fän

Dang

13:09

Good ol' MacBook Pro with US-based keyboard (I hate it)

if learath is on pc he is probs on gentoo or complaining about systemd arch

Fän

Good ol' MacBook Pro with US-based keyboard (I hate it)

The keyboard or the macbook or both?

The keyboard

13:10

the macbook is ok, but 60hz really sucks

Ryozuki

if learath is on pc he is probs on gentoo or complaining about systemd arch

Dont get me started on systemd PaimonAngeryWOKE

PaimonAngeryWOKE

Embedded image

Fän

The keyboard

Is it the butterfly switches that annoy you or the layout?

today ill get my ryzen 5600x, my rust programs will compile faster monkalaugh

monkalaugh

Learath2

Is it the butterfly switches that annoy you or the layout?

I'm EU-based (surprise surprise), and it's a US keyboard. See why I'm annoyed yet?

I had to get the damn keyboard replaced 5 times because the butterfly switches are awful

Ryozuki

today ill get my ryzen 5600x, my rust programs will compile faster monkalaugh

monkalaugh

what did you have?

ryzen 2600

bruh xd

xd

that is surely old

13:11

lol

works wonders tho

switch to am5 kekw

It's a great entry CPU for sure

2600 came out in 2018

13:12

xd

13:12

not that old

gerdoe

switch to am5 kekw

boo ddr5 booo

Fän

I'm EU-based (surprise surprise), and it's a US keyboard. See why I'm annoyed yet?

Huh, I love US keyboards. Though I also just replace all accented characters with non accented ones. Makes me look uneducated but whatever, I'd rather have {}[]`~ easily accessible :P

I code on QWERTZ keyboard, but my MacBook is QWERTY. I always need to get used to whatever keyboard I'm working on at that moment

13:14

And don't even get me started about how "@" is ctrl + L on the macbook (??????) (edited)

I have karabiner on my mac so I can make the keymap match my desktop keyboard

13:15

Have you tried just remapping a bit?

Meh

Hm @ is shift 2 for me

alt gr 2

My huge issue here is at uni, the computers all have italian keyboards

Makes me code at snail pace. I look like an old man trying to find keys

Also let's not talk about the useless touch bar at the top which is really only useful for emojis (which aren't even useful)

Yeah, I miss my physical f keys and esc

I can't even ctrl + f5 to reload cache.

Having a touch esc is so annoying in emacs and vim

13:18

I mapped esc to `

your ESC key is touch? o.O

Yeah the shitty butterfly keyboard also came with that

https://gyazo.com/b483eb2d1a51f2c9d73500c829bdebd2

Image attachment

this is what my touch bar looks like (edited)

13:20

Though this is a M1 macbook

Yeah touch esc was such a horrible idea they changed it the next gen, they also replaced the horrible butterfly switches

maybe older mac's have touch ESC

13:20

oh yeah

13:20

I can imagine how that is awful

Oh, if you have an m1 mac can you check if ddnet works there?

Learath2

My huge issue here is at uni, the computers all have italian keyboards

makes you write spaghetti code monkalaugh

monkalaugh

monkalaugh

4

bluekitty

1

Learath2

Oh, if you have an m1 mac can you check if ddnet works there?

This M1 Macbook Pro is the first Apple device I bought, ever (edited)

13:22

that's why I was so surprised this is only 60hz

Ravie

makes you write spaghetti code monkalaugh

monkalaugh

anyway yes I can check

13:23

let's see lol

I like the idea of M1 a lot. Hopefully other vendors will get on the hybrid core trend soon

for some reason i read iMac there

13:25

works fine on m1

Great \○/ don't need to bother with an arm build

Learath2

Or if this is only deployed during attacks maybe we should just not advertise on the 0.7 master during that time

Why?

People wouldn’t be able to join if the filter is set up as I think it is

They can't as of right now

13:42

again, easy to implement once the new stuff arrives in frankfurt

So registering on the 0.7 master would be wrong

Well, 0.7 can be verified just by a refresh of the server list, because the serverinfo requests uses tokens already

13:43

And 0.7 still pings every server manually

Saiborg

What is the reason we cant reconnect unlimited times to a server and we get timed out?

Ive previously asked this question, but had to unfortunalty leave before i could read the answers... basically i know a way to bypass the "maximum reconnect" attempts, and have been using it when i want to get into a server that is. Its a nice thing, but at the same time if its a security risk, its better that it gets patched

I doubt you are actually bypassing the limit

17:17

unless you are changing ips

pog (@Learath2)

We'll probably sooner or later move to 0.7 protocol ourselves

@Saiborg I think there's no limit to reconnect attempts, the server just bans if you reconnect too fast

20:53

if you still think you've found a bug, feel free to PM Learath2 or me

20:53

@Fän do you have filters for existing UDP games?

20:53

(sorry for all the protocol replacement talk, I'm sure you're getting annoyed at it already given that you deployed filters for the current protocol)

20:54

which protocol would we have to imitate to get the best DoS provider coverage?

https://drewdevault.com/2020/09/25/A-story-of-two-libcs.html

21:33

(why isalnum can segfault with glibc)

localization makes everything worse

for the programmer

the users also don't enjoy the increased complexity leading to bugs and crashes

my parents wouldn't enjoy english software

Sure, it can be a requirement, but sometimes it's overused

Exported 479 message(s)