[DISC/RANT]Why no support for peer to peer?

Tomsamson: my post was simply refuting your comment that a RIA without flash would "probably suck" so I gave multiple examples of well-regarded RIA examples that don't use Flash.

Additionally, you are making a HUGE assumption that it takes longer to developer RIA apps using technologies other then Flash. This has neither been proven nor disproven. I really question how much more productive it is at times. The learning curve for Flash is pretty extensive in it's own right and while some things are very easy, Flash makes other things quite a bit harder (injecting dynamic content site-wide for instance, site templates are another good one).

Quote:

---

Originally Posted by webgeek

you are making a HUGE assumption that ..

---

nay,it was no assumption,as i said i spoke about my own personal view and experience after working several years in a web agency. Your experience may vary/be different.
Also,sure,flash is better suited for some things than others,again,i didn´t say skip all other technologies/languages,i was more into arguing for using each for the part where it makes sense.
Somehow all flash developers i got to know were more open to using flash together with something else or not using flash where something else makes more sense than non flash developers are in some cases open to use flash where using flash makes more sense,that´s why i defended flash´s position some,that´s all :)
I just had such discussions often,in most cases with js developers who saw flash as the all evil thing but the next day moaned again that something (made with CSS/html/js only) looks way different in IE than FF :)

I still don't understand much about computers, may I ask a stupid question? :)

What kind of application would someone like to make when flash does support p2p?

One could create instant messaging or file sharing applications and generally anything where collaboration/networking of several machines is wanted (with the advantage that the "performance" (bandwith,proccessing power,storage capabilities etc) of the network raises as the amount of machines taking part raises (contrary to what happens in client/server model)

P2P is used in all the file sharing networks. In this case, I assume it would be used for Flash games generally. Basically client A starts a game and client B connects to client A's flash player. This direct connection allows for faster data transfer as it doesn't involve an intermediate server. Basically, one of the clients is running as the server for the game as well as a client.

Aha, I see.

multiplayer-games ...
I'd rather have 3d-support first in that case :)

- Dang 10 minute edit, this was part of my last post but I took too long to post the edit.

Quote:

---

(with the advantage that the "performance" of the network raises as the amount of machines taking part raises (contrary to what happens in client/server model)

---

This isn't necessarily true. For instance, if there are three players you can design it three ways:

P2P Everyone connected to everything
a <--> b
a <--> c
b <--> c

Client-server
server <--> a
server <--> b
server <--> c

P2P distributed
a <--> b
b <--> c

In the first case everyone is interconnected, if "b" does something that "a" and "c" needs to see, it will be forced to send a message to both "a" and "c" directly. That's two outbound messages from the client.

In the second case "a", "b", and "c" are connected to the server. If "b" needs to send a message to "a" and "c" then it just sends it to the server who then sends it on to "c" and internally routes to "a". This involved an extra hop to get to "a", and "c" BUT the client originating the message sent only one update. Additionally, it's a consistent single extra hop. No routing needed beyond the initial delivery.

In the third case, P2P network style, "a" is connected to "b", and "b" is connected to "c". For "a" to send a message to "c" it needs to route through "b". This means that there is an unnecessary hop AND "b" is wasting bandwidth for something it doesn't even care about. Ultimately this is very ineffecient as the number of hops between clients grows.

-=-=-

The first model seems good but becomes impractical very quickly. You can only have so many connections before there are issues. Additionally, you might be broadcasting a single message many times to reach everyone needed thus burning your own bandwidth. Finally, in the "everyone to everything" model there are practical issues. If this was a game, which machine controls the "game state"? And how do you prevent cheaters, they could modify their client to give them unfair advantages. I didn't make either of those problems up, both have been seen in commercial games.

The third model is how you initially search a file sharing network, which is why it takes so long to get results back from something like e-mule. Then, when you start transfering data, you fall into the first model where a direct connection between clients is established. This is fast because there are the fewest number of hops. Additionally as there is no centralized server, it can't be shut down easily (remember Napster in the old days?).

In general the client-server produces the least amount of "waste" bandwidth as the client sends only what it must and the server sends only what has to be sent to each client. No one gets messages they don't need or care about (unlike the routing in P2P). Security is easy as well, the server can be hardened and validate messages to prevent hacking. There is a reason this model is used for almost all multiplayer games. The down-side is that the server has to have the capacity to support the entire network (both bandwidth and CPU/memory, etc).

SO... Back to the quote... P2P (either distributed or everyone to everything) provides the highest amount of overall bandwidth as you are sharing everyone elses. It can also provide the worst latency (by a huge margin) due to the number of hops. So performance might or might not be better depending on what you are doing. In generally, true P2P doesn't buy much in most cases. It's excellent for file sharing and instant messaging (although it's not used for IM all the time either, many are client-server) but not for a lot of other things.

I can see where P2P may be useful, but it seems that it would add a lot of complexity as well as making it incredibly difficult to prevent hacking.

Webgeek: Interesting post :)

Since we're talking hypothetically "if" flash had p2p, here's what I concluded would be the ideal setup based on what you've posted above:

Regardless of what method you chose to distribute actual game data, you need a server side app, if only to connect clients to each other, keep track of games and perhaps provide centralized statistic gathering.

Once a game starts one client is chosen to act as the server. All other players connect to this one player who broadcasts all relevant information to everyone. To choose who acts as the server in game, the central server would do some checks (eg: have each client ping everyone, then choose the client best average latency)

The game would also need to be able to select a new client to act as a server on the fly. So say the client who's acting as server drops out, loses the game or too many clients report that it's latency is too high the central server can broadcast to each client who the new server is, whom they reconnect with and continue with the game.

This obviously wouldn't work for some game types, especially anything realtime where a low latency will provide advantage since the server client will essentially have 0 latency. For turnbased games, or realtime instances where latency isn't an issue this setup would ideal.

Of course, you've still got the problem of the flash player and how incredibly easy it would be for the server to cheat... Not a great deal you could do about that though I guess.

yeah, good post webgeek, i´m way less a mp gaming guy than you are,it was interesting to hear in which cases p2p could act worse.
Speaking in general it seems to me like the best setup could be a mix of the upsides of both client/server and p2p,like there would be a server to init and maintain global things,also broadcast things when something has to be broadcasted to all or to unspecified client(s) but once everything is set up and its about communication between two set clients one would switch to something like p2p :)
I didn´t even think about using p2p for mp gaming but now i´m curious, webgeek,would that be a setup you´d think would be good suited for mp gaming? (besides the point parts of it are not supported with flash player :) )

Honestly, it depends very much on the game itself. Most MP games are client-server. All MMOs are that way and anything like Battle Net or the various services like that are client-server. Quake and the associated FPS games are client-server as well.

I know some of the RTS games used to be client-server but one client acted as the server for everyone. I'm not sure if they still do this though because there were issues with "patched" clients that would show the whole map and all your opponents moves as long as you were running the server on your machine.

So in general, even with P2P available, almost all games are client-server. At the minimum one of the clients will be running the server interally.

i see,well,still interesting,i think there could be some shift towards a more mixed integration of both in the future :)
Anyway, for the meantime i´m fine with what flash allows regarding mp gaming right now, looking forward to actually max that out with a fitting project before asking for p2p or a mixed combo setup :)

Quote:

---

Originally Posted by tomsamson

My colleagues would just slowly and step by step accept flash wasn´t a kiddy scripter toy anymore but got a more and more mature language and you could do things in it which wouldn´t fly in some other web languages at all or don´t act the same accross systems at all.
People who are still heavilly and generally against flash should wake up and realize that in this day and age someone who has flash playback blocked in his browser probably has a similar crippled experience as if he had js turned off,its just used for that many things.

---

This is what I was talking about. Flash is mature and stable now, new versions of the player and the IDE come out every so often and are quickly adopted. It is like a lightweight java now. It's a quality product and people know that. Why not add a feature or two every now and then instead of worrying about exploits and whatnot. People can choose whether to download it or not. I don't think people are going to snub it just because it can connect to computers other than the site it came from or because the plugin/player is 5 megs instead of 2.

i still think one has to differentiate between adding filesize of plugin raising features and ones which besides that probably also add security exploits.
Personally i´m for getting 3d or generally speaking hardware acceleration features added to speed up the display/rendering side,cause the worst thing that could happen (which i think adobe could avoid pretty well) would be unstable playback performance on different system setups.
Reducing restrictions of the security sandbox on the other hand can have much more/bigger downsides (which can cause constant damage to hardware/software in indirect manner) so i think its good adobe is careful on that end (i wouldn´t like it if the audience of the flash plugin decreases dramatically cause it gets known it has a big security issue which won´t get resolved)

It's time to resuscitate this thread because P2P will be supported very soon by FP10 + Stratus

http://labs.adobe.com/wiki/index.php/Stratus

( thx to Tony for his post on mochiads forums
https://www.mochiads.com/community/f...ke-me-a-server )

Quote:

---

Originally Posted by Sietjp

It's time to resuscitate this thread because P2P will be supported very soon by FP10 + Stratus

http://labs.adobe.com/wiki/index.php/Stratus

( thx to Tony for his post on mochiads forums
https://www.mochiads.com/community/f...ke-me-a-server )

---

Wow.... both P2P and UDP! My question though, is what happens after the beta ends? They didn't mention anything about that in their FAQ. Is it going to be an expensive service we have to pay for? Are we going to have Stratus available to use on our own servers?

Perhaps I am misunderstanding the idea behind Stratus but it is not piece of software you can install on your own server. Its a service provided by Adobe that your swf can connect to and is probably running something like Adobe Media Server which could be set up in your own server too, only its overpriced and so complicated that only people who understand how it works and how to set it up properly are the part of Adobe development team.

There has been no mentioning of prices. Knowing Adobe, I would not be surprised if the service would end up being free in US, costing hefty sums in Europe and blocking every user from other parts of the world like Russia or China because of security problems.