i created my own protocol for my games...

Рет қаралды 67,217

Күн бұрын

Пікірлер: 151

@notrealsaga 3 ай бұрын

Watching people who are great at using vim is like watching people training at the gym: you find them sexy, you want to be like them, but you don't want to go through the effort

@Skaffa 3 ай бұрын

i like the way you worded that

@Kane0123 3 ай бұрын

This is perfeft

@moonasha 3 ай бұрын

not even worth the effort for me. I spend maybe 5% of the time actually coding, the rest is switching between files, peeking at classes, debugging. All the shortcuts don't mean diddly squat for me. But visual studio's tab layout and ability to peek at stuff really quickly helps.

@hamm8934 3 ай бұрын

Vim speeds up switching files, peeking at classes, and debugging, and much more. Vim isnt just about the improved speed writing code. Vim speeds up window management, navigating files, moving window panes, etc..

@AG-ur1lj 3 ай бұрын

@@moonashaI have vim shortcuts for all those things… project-wide grep, pinned files to jump to, jump to definition, jump to next reference, all of it

@Crcs-1997 3 ай бұрын

These video formats are my favorite. Love the content

@kaibe5241 3 ай бұрын

Man, really wanna see more of these kinds of videos - your deep dive into this, showcasing also your knowledge of things like TCP protocol and websockets, is super interesting :)

@AlucardNoir 3 ай бұрын

"[...]I think that's probably more than enough, than I need" You're just asking for trouble now.

@Exilum 3 ай бұрын

That's what the version tag is for

@Moner-nt5td 3 ай бұрын

I created my protocol 10 years ago, consider a short in the packet for routing, do not use a int for len unless you want recieve 2 gb packets, handle all connections and packets in a own async thread pool. the CPU has a fast array copy instruction, keep that in mind.

@TheScabbage 3 ай бұрын

I've been using little-endian in my protocols for years, just because all the major architectures use it. It simplifies a lot of code and saves the small perf hit of having to convert back and forth constantly.

@HobokerDev 3 ай бұрын

Prime leveling his scope creep skills.

@kefpull6676 3 ай бұрын

"kinda FUN" (laughs in 9999 possible vulnerabilities)

@AstuteJoe 3 ай бұрын

Kinda the only way to make profitable realtime* multiplayer games unfortunately though

@ahmednishaal9432 3 ай бұрын

Let's have Thor ethically hack his servers 😂😂😂

@manuel_g_g 3 ай бұрын

first make shit work, then make it safe. Securing a protocol/channel is not an easy task, and you probably need the support of your Russian/Asian hacker friend to do it.

@AstuteJoe 3 ай бұрын

@@manuel_g_g Not quite the case with multiplayer games too, if you don't plan the architecture ahead, and consider the gameplay implications of each architecture, you can end up with a networking model that will never be cheat proof, like making a peer to peer competitive shooter

@manuel_g_g 3 ай бұрын

@@AstuteJoe you have a point. And I think even now, hundreds of AAA game's middleware engineers must be working on it constantly to find an effective way...but the nature of the domain is what should be addressed on a case-by-case basis.

@RedSpark_ 3 ай бұрын

I love these videos, they're so inspiring. I started building a text based adventure game to learn about OOP and some design patterns, but now I'm thinking about how I could make it multiplayer.

@AJMansfield1 3 ай бұрын

I've built my own network format, it was horrible to debug because it was specifically designed to resist casual inspection -- using a hash-based addressing scheme where each station needed to use its cryptographic key to decide whether or not any given packet was addressed to and/or meant to be relayed by them.

@kernel0verflow919 3 ай бұрын

Even though it’s some people thinks is depressing to see talented people code something, for me it’s motivating to work on my own skills and become better in what I’m doing. Thanks for the videos and sharing your progress prime 😇

@privacyvalued4134 3 ай бұрын

0:29 I've made so many protocols at this point, from binary to text protocols, that this is child's play for me. My recommendation for anyone's first network protocol: Two-way JSON separated by newlines. Is it efficient? Nope. JSON encoding/decoding performance is always terrible. Will it work surprisingly well? Yup! Can you easily interop fairly complex data across multiple languages and avoid a ton of bugs and security pitfalls? Sure. Can you alter it without worrying about its structure? Absolutely.

@nexovec 3 ай бұрын

Yes, I did. Unfortunately it was written in lua, so as you can imagine, the moment I stopped sending string identifiers was the moment it kind of lost its ground.

@uuu12343 3 ай бұрын

Everytime I see prime create something, I'm torn between watching or not watching the video because watching the video would risk me completely dying inside from the sheer impending feeling of being a failure and imposter hit

@TehKarmalizer 3 ай бұрын

Why? Did his wife ask you to murder him and assume his identity?

@sskeptix 3 ай бұрын

I played with that stuff 2 years ago. I did almost the same except I didn't have an encoding flag but instead I had a packageId flag. You may have a situation when your client sends multiple packages and the server can spend a different amount of time handling them. Client sample sequence of events: Send1, send2, send3, receive response1, receive3, send4, receive2, receive4 You may have 2 types of packages: notifications (fire and forget), request, and response.

@GlennLewis 3 ай бұрын

Remember the Amiga IFF ("Interchange File Format")? This brings back lots of fond memories. 🙂 Thanks for sharing!

@Patterner 3 ай бұрын

sometimes in my nightmares after i wake up screaming

@KvapuJanjalia 3 ай бұрын

I've created my own protocol but with variable length integers in header, because it is more _fun._

@pixelfingers 3 ай бұрын

Love these videos, you’re very good at describing and teaching, excellent ❤

@mickey20512 3 ай бұрын

More of these videos please! love these insights and learn heaps more then the react vidoes

@valentinrafael9201 3 ай бұрын

Dude, you are inspiring!

@regiondeltas 3 ай бұрын

This is my cup of tea. I'm not a developer by trade but I've been working on a Pico powered hobby project, in rust because of course. I've written an extensive low level TCP protocol for data transfer, super lightweight, extensible, variable packet sizes and data types etc and with a nice shared library so that I can use the same core code for both the server (again, in rust) and the pico client. You learn a LOT when implementing things at this level and its so satisfying when you get it working end to end. Got a simple UDP discovery protocol too, which was fun to do

@vessbakalov8958 3 ай бұрын

This was really good.

@ericmackrodt9441 3 ай бұрын

That's very similar to the Napster protocol. I made a similar one to communicate with a Windows 3.11 chat app I'm building. The difference is that the actual values are sent in a specific order separated by spaces after the length.

@wdavid3116 3 ай бұрын

If the packet is being sent via UDP you don't have to send the length because the UDP packet has a length associated with it and it will be received as a discrete message. If you send the packet via TCP (unlikely to be a good idea for real time game data,) you absolutely need a length to know where in the stream a particular message ends, or you have to do something ugly like a delimiter or fixed size messages or something. There isn't really any reason to use bigendian. Virtually all modern CPUs use little endian though it is pretty trivial to byteswap from big endian. The only potential advantage is that reading the data in hex format will be easier with bigendian but there isn't any reason to do that when you developed the protocol and clearly have code to parse it and present it to you in a readable fashion. It isn't very wasteful but unless you're doing raw sockets or something there is no reason to do big endian. Personally I really wish go had a c struct compatibility library similar to the python struct library. That would take care of all the padding and you could just use structures fill them with data and send them through sockets without serialising anything (as long as all the struct members are simple data types or structs composed of them with no pointers.) I think you can do this with unsafe go but then you lose a bunch of the benefits of using go. I settled on just using protobufs for my protocols in go. They aren't for a game but likely would be just fine for basically any game someone might write and I can generate code for all the other languages I use without doing anything. I almost went with CapNProto but it didn't support all the languages I jump between for my side projects.

@kajika135bis 3 ай бұрын

"if count == 5" should be "if count == BUF_COUNT", just sayin

@adriancruz2822 3 ай бұрын

All my homies hate magic numbers!

@_Gart_ 3 ай бұрын

I need the matrix so I can download programming knowledge like this guy. He's so fast

@hidoryy 3 ай бұрын

what are the benefits of asserting in runtime rather than in unit tests? actually never seen assertions being used outside of test scope

@Jamo008 3 ай бұрын

The last screen Run fund, WithCancel is a noop cancel isn’t cancelling outer

@ja31ya 3 ай бұрын

When your content is over 9000 but your volume is -9000

@simplyblunder 3 ай бұрын

Loving seeing prime using GO!

@barry_wastaken 3 ай бұрын

Wait untill he finds out about event driven architectures, he will have no hair left.

@metaltyphoon 3 ай бұрын

Lol with those asserts and break the world all it takes is non ECC memory to fail in the right place and it all falls apart.

@ra-dro 3 ай бұрын

Amazing video!

@pdougall1 3 ай бұрын

Such a great experience in having a horrible time... my whole existence

@pedrosilva1437 3 ай бұрын

Back in the early to mid 90s, ASSERTs in C/C++ were a very common thing. It seemed to go out of style... we may have been smarter back then. 😉

@AshesWake-sf7uw 3 ай бұрын

Man i wasn't even born back then, unfair!

@masterflitzer 3 ай бұрын

11:32 wait prime is not raw dogging git cli?

@ErazerPT 3 ай бұрын

Kinda wondering if just having packet[2]+packet[3]*256 (or packet[3]+packet[2]*256 depending on how you wrote it) would have any significant speed difference over a function call that basically does... just that. An optimizing compiler should inline it, but...

@cennarr 3 ай бұрын

How did The Primeagen make that ascii diagram? Any tool he uses for that? Because it seems like hell to draw all those borders and such by hand

@bautistarescala 3 ай бұрын

May i know how you type those sequence diagrams easily? I assume it's some Vim macro but i don't use Vim so i don't know but i'd like to learn. Seems pretty useful.

@reybontje2375 3 ай бұрын

Fixed-length encoding for the length header. Gigachad. I don't know if it was a conscious choice, but using variable-length integers for the length header drastically increases the complexity of parsing the header in a safe/secure way that doesn't open you up to DoS attacks that come from carefully engineered messages.

@aredrih6723 3 ай бұрын

The assert around 7:14 checking that the packet length is greater than 0 is a DoS waiting to happen (a message with no body will crash the server). As for using variable length integers in headers, beside overflows in naive parsing, I fail to see the issue with them. I agree more considerations and processing are required to handle variable length fields and their flexibility is really too useful but IMHO the difference between fixed size and variable size isn't that large and you can reuse the code elsewhere. Unless you plan to decode raw messages by hand (in this case, a decoding util is pretty good), I fail to see the difference.

@AshesWake-sf7uw 3 ай бұрын

@@aredrih6723 Can you pls explain how is that vulnerable to DoS?

@aredrih6723 3 ай бұрын

@@AshesWake-sf7uw an assert failing is a program crash. If Prime goes "tiger style" (my understanding is that he is a fan of the work on tiger beetle, so likely) and includes the assert in the production code, a packet received triggering the assert will bring down the app. DoS is denial of service. Crashing an app counts as a denial of service. An "exploit" would be waiting for the server to start, open a TCP connection, send a packet with an empty body, and the assert would kill the server. It's a 10 seconds fix so not a big issue but still classify as DoS.

@moaxcp 3 ай бұрын

Why don't you set up the version and encoding when the connection starts rather than in each message sent?

@lovesowing 3 ай бұрын

when did prime switch to jetbrains mono? I thought he didn't care about fonts. next thing we know he'll be configuring hyprland

@gamemusicmeltingpot2192 3 ай бұрын

I worked on a game that had this type of special protocol

@niklaswojtkowiak3 3 ай бұрын

Was this supposed to be on the Vimagen?

@MrAbrazildo 3 ай бұрын

4:05, if PACKET_ ... _SIZE is a constant, shouldn't this assert be 'n < PACKET...'? 4:08, literal numbers... the most underestimated issue of all time! They should be MAX_TYPE_SIZE and ENCOD_SHIFT. 6:36, couldn't it be bigger? In this case, your if would fail.

@ricky2629 3 ай бұрын

Am i missin something or wouldn't that assert for lenght of the message crash the server on any malformed request?

@kimberlycelaya1895 3 ай бұрын

Welcome💙 🔥🔥🔥

@TheDoppelganger29 3 ай бұрын

Just where do you find time after work??

@johnyewtube2286 3 ай бұрын

He no longer works at Netflix, he is full time e celeb.

@portal-jx5pu 3 ай бұрын

Nice protocol

@jop0693 3 ай бұрын

Anyone knows how he did the markdown graphics?

@digiryde 3 ай бұрын

As to creating your own packet structure.... Do you have to have succeeded for it to count?

@onegraham 3 ай бұрын

Why do you need so many bits for version?

@kengreeff 3 ай бұрын

It definitely is an idea 😅

@dennisloska 3 ай бұрын

What's the fonts/plugin name which Prime is using here? The != and

@remains2246 3 ай бұрын

they are ligatures

@TheHeavenlyDemonSupreme 3 ай бұрын

The font is JetBrainsMono and those are ligatures as another commenter already mentioned. I'm not sure but I think this is the default font for Ghostty. It doesn't look like Prime has done much if any customization.

@dennisloska 3 ай бұрын

@@remains2246 I learned somethong new today, thanks man!

@dennisloska 3 ай бұрын

@@TheHeavenlyDemonSupreme Thanks!

@bytebytecode 3 ай бұрын

I didn't understand anything, but it was fun watching 😅

@hydrobolix3365 3 ай бұрын

🤘

@AndreanFranc 3 ай бұрын

I think little-endian makes more sense (even though it's not the network byte order), just because most of the architectures are little-endian now anyway. Better pay a penalty for conversion only on the outlier architectures (big-endian, if they even exist as your clients), than pay for it everywhere in a wasteful manner.

@georgehelyar 3 ай бұрын

I've done version, flags, opcode, data, where data might be length prefixed but it depends on the opcode whether it's variable length. Big endian is not really needed at the application level, you can just specify the endianness in the protocol. Protobuf pretty much means that you don't need to write your own protocol any more though, as it already does variable length numbers efficiently, backwards compatibility etc. You can save a few bytes here or there with a custom protocol but it's usually not worth it. Also multipart is surprisingly good if you want to do human readable, but most of the time you don't want to do human readable.

@HypeLevels 3 ай бұрын

ligmuh (I am scared let's watch this 😂)

@ոakedsquirtle 3 ай бұрын

okay but what is a framer

@Noober666 3 ай бұрын

I already knew how to do this, just wasn’t gonna tell u 😏

@luizalbertolausdarosa6819 3 ай бұрын

Now do it with UDP, please!

@justanothercomment416 3 ай бұрын

Basically the same thing. He can either fill a buffer with multiple packets and send that as a datagram (looping on read), or simply send/recv on packet per datagram. TCP is usually slightly more involved because you need to know how to parse the stream and to account for short reads and so on. Which requires lengths so as to know when the full packet (or plural) has arrived. He has lengths embedded in his packets so it's good.

@memory_leaps 3 ай бұрын

how did u made the ascii tables

@ItzD3fW1sH 3 ай бұрын

Plenty of online generators for it

@codingwithifeanyi5379 3 ай бұрын

it this go?

@bertram-raven 3 ай бұрын

Naive in so many ways from the start. Does it get better? Where is the error correction? Where is the message resync?

@JohnLovell-FTW 3 ай бұрын

You didn't use Mermaid in your Markdown? Shame! :)

@ashleigh. 3 ай бұрын

Please, if you're going to do this, use QUIC

@Cdaprod 3 ай бұрын

All the time.

@aajas 3 ай бұрын

6c is lowercase, the thumbnail has uppercase L

@ThePrimeTimeagen 3 ай бұрын

such facts, well observed

@milseq 3 ай бұрын

Go is so damn beautiful I really need to learn it and find a project to use it on

@asmod4n 3 ай бұрын

you can send empty packets for keep alive :)

@greyshopleskin2315 3 ай бұрын

The kernel should automatically send keep alive packets (assuming the socket type is tcp)

@asmod4n 3 ай бұрын

Realtime communication protocols aren't tcp. I also believe you can restrict usage of tcp keep alive for apps on some operating systems.

@canofpulp 3 ай бұрын

Microsoft recall is enabled by default in the latest Windows update!!!!!!!!! update 24H2

@Mraeth8n 3 ай бұрын

why did you choose 8 bits for the version? Wouldn't like maybe even 4 be sufficient?

@chinoto1 3 ай бұрын

If you're careful, but if you run fast and break things constantly... Maybe he should have gone bigger 🤣

@Mraeth8n 3 ай бұрын

@@chinoto1 haha 🤣

@NostraDavid2 3 ай бұрын

What? You've got a nice encoding-decoding setup going and you're NOT PBTing that shit? HAWK TUA ON THAT CODE! Slapping some Property-Based Testing on code that bijective code is the best you can do here!