Hello everyone,
i’ve been lurking for a while and now it’s time for me to post some questions i’m thinking about.
I’m working for a small italian company who is developing a small client-server application in which users will be able to chat and use/watch their own multimedia contents, like images, audios and videos.
All this stuff will be stored on a central server, and uploaded through a web page.
Now i’m thinking/shaping the networking part of the code which basicly have to do a user/pass login, keep up a small chat system, retrieve the list of available contents for that user and, when requested, retrieve the content.
Now teh questions:
login security: can i use https with wwwform? Can RPC calls use any sort of security algorithm (like sha1)? Between wwwform and rpcs, which way is better/more secure for this task?
I’d like to download contents and lists directly through www() calls, thus bypassing Game Server, to speed up the whole thing. Is this safe?
Game Server is going to be a sort of relay (receive from one, sends to all) for player’s positions and chat messages. Do i need to write my own ultra-optimized-raknet-based game server or can i use something built in unity? Where do i find info on how ot make the raknet-based one? What is (more or less) the cap to the number of users in each way?
I’d like to use javascript to build all this procedures. Am i creating a bottleneck because of js interpretation? Would c# be way better?
Concerning HTTPS and WWWform: I think I’ve read that some people had trouble with this, but I haven’t tried it myself. Could have been just an invalid certificate or something like that. Some pointers that might shed light on this:
I guess this depends on whether the WWW class does support HTTPS or not. If it does, it should be safe, if not … you might have to implement your own security layer to make it safe. Or talk to UT to see if there’s a way to make HTTPS work with the WWW class (if it really doesn’t support this, yet, this is something I would consider a bug ).
From the “networking perspective” (i.e. stability), it should be safe. In other words: WWW access should not interfere with Unity networking.
From your description, it seems that the application is simple enough that you should be able to build a “Unity server”. The only limitation I currently see with that approach is that while there seems to be a headless server mode on Mac, there’s no such thing (yet) for Windows, and Windows server hosting is usually cheaper and easier than Mac server hosting.
Since there’s no heavy game logic involved, you should be able to host many more users than for any typical multiplayer game, but I guess the only way to get reliable numbers is creating a simple testing simulation and testing as many users as you feel you need.
You could either implement your own position synchronization via RPCs, or use the NetworkView synchronization. In both cases, it should be possible to create some sort of simulation that will spit out somewhat reliable numbers.
With my current game (Traces of Illumination, previously TRaceON), I can simulate around 10 players on a single machine (Mac Pro) with the actual game clients (“AI controlled”) - but that’s a “real game”, so I guess with something much simpler (and more designed for simulating heavy loads of users), you could simulate many more users on a single machine, so you might be able to simulate intense numbers of users with just a couple of machines.
One thing to keep in mind if you’re creating a Unity server: Make sure everything that’s not absolutely needed is switched off on the server. The nice thing about this approach is that you have just a single project which ends up being both client and server (with conditional compilation, where needed, if needed). But obviously, you don’t want to render everything on a server that no one looks at
Oh oh, JavaScript vs C#
But seriously: If JavaScript is your preferred language, you should go with that. In Unity, both C# and JavaScript are compiled into bytecode and then run in a virtual machine - so at runtime, there should not be any performance differences, except for when you use JavaScript features that consume time (dynamic typing). There’s something like @pragma strict to prevent usage of such features (but I’m not a JavaScript guy, so this is just from faint memory ).
Yes. It somewhere says that this is supported now in pro (this posting, to be exact), but I’ve just tried that with my game server, on my Mac, using the approach as described in this posting, and while I do get a message “Starting Server”, I can’t find the process, and I also don’t get any logging (which I’m doing via log4net, so this should go into the relevant file).
When I start the server “normally”, i.e. with GUI, everything’s working just fine.
So, I really wish there was some documentation on this -batchmode feature
