Fly Free?!

[legoboyvdlp]

How many people on wildly undersized machines will press that button and then cry about FG makes their box slow?

Don't press buttons if you don't have any idea of what they are going to do!

As one young air force pilot was once told * "Never touch any buttons, or pull any levers unless you are told to."
"Yes, sir"
In under an hour he was in hospital due to injuries sustained when doing an unofficial "test" of the ejection seat.
When he was back to flying duty, he refused to even push the throttle open on his first solo until he had direct orders to do it.

*In my imagination

[Jabberwocky]

@wibragg:

Don't press buttons if you don't have any idea of what they are going to do!

Now, isn't that true for anything that has to do with planes?
But seriously, since, according to Albert Einstein only two infinite thigs exist, the universe and human stupidity and he was never sure about the universe, ... well, there is a logical conclusion why not to do things like adding servers to a client package.

I looked a little bit into HLA, but not enough to understand all the consequences yet. I started some of my own experiments, not so much with the goal to build an own MP structure, more with trackers, crew-intercom and maybe some tactical data exchange in mind via http. But due to other work in real life, I never made it beyond a basic prove of concept. So, some work there which is independent from the general MP structure/HLA.
The problem with HLA as I understand it, is, it is exactly that, a High Level Architecture. It basically defines objects (for example a plane or a runway) and relations and interactions between them (as in plane land on runway). That is of course very simplified, but it shows already the basic problem. It is high-level, means, machine independent, just another level on top of the already existing ones. Which means more coding and decoding information at runtime. I doubt, an already slow and overburdened system will go faster by putting MORE work on it.
So, HLA will surely bring some advantages, but as I see it, ONLY if the underlaying levels are cleaned out and streamlined first.

J.

[Hooray]

It is high-level, means, machine independent, just another level on top of the already existing ones. Which means more coding and decoding information at runtime. I doubt, an already slow and overburdened system will go faster by putting MORE work on it.
So, HLA will surely bring some advantages, but as I see it, ONLY if the underlaying levels are cleaned out and streamlined first.

Sorry, but you may want to read up on HLA once again - you simply misunderstand the relationship and advantages when comparing the existing MP protocol with using HLA. There's plenty of information on the wiki about our existing MP protocol (and some on HLA) - but basically it works in a "polling" fashion, while HLA -among a plethora of other benefits- provides PUB/SUB support: https://en.wikipedia.org/wiki/Publish%E ... be_pattern

Thus, even a simple HLA-based implementation among a handful of fgfs clients is likely to outperform the existing MP protocol in terms of bandwidth/performance considerations.
And like you said, HLA is much "smarter" by dealing with objects/entities - so it can provide custom message types per entity/object, which can tremendously streamline communications to update state in a selective fashion.

For instance, imagine synchronizing AI state, weather/environment state or other simulator state in an MP environment - you could even propagate entirely custom events (think aircraft/combbat damage) or skid marks across all clients easily.

Obviously, you can also do all of this using the existing protocol, but under the hood, this is what the MP protocol looks like:


with HLA being this:


The existing system happens to work well enough for most people (and is robust, simple and easily understood), but only because of their low requirements - people wanting to do more involved things will sooner or later want to use features provided by HLA.

[wlbragg]

Hooray, your pictures speak a 100 1000 words.

I doubt, an already slow and overburdened system will go faster by putting MORE work on it

This concept, or should I say the underlying cause, using older hardware, is prevalent in this community and I get it. But it is counter productive to stop the progress of a project like this because of dinosaur DNA. I understand there is line there somewhere between tin can and water cooled superconductor.

[Johan G]

I am not that sure HLA is all that performance hungry, considering that it was already used (experimentally) in distributed (as in interconnected) real time simulator environments already in the mid to late 1990's. Surely both hardware and bandwidth, even if nearly "antiquated", would outperform even research grade systems used then.

[Jabberwocky]

The point in question is not HLA. You don't need to convince me, I am already convinced. But the point, I am thinking about is how the HLA data is transmitted. As Hooray wrote ...

Thus, even a simple HLA-based implementation among a handful of fgfs clients is likely to outperform the existing MP protocol in terms of bandwidth/performance considerations.

, this is the point. See. if it would be direct between clients, all would be good, but then every client has to send his information ALL the other clients. And since every one does that ...

2 clients -> each has to send every message 1 time -> 2 messages total
3 clients -> each has to send every message 2 times -> 6 messages total
4 clients -> each has to send every message 3 times -> 12 messages in total

so bilateral communication is obviously impossible as soon as the number of clients grows. Which brings us back to communication with servers in the middle. Which well, the server gets all the messages and has to send them to all the clients to get the data in their HLA object buffer. Unless they would actually use a bean-like client-server-shared object to represent the HLA objects. And if that part is messed up, you wonderful iPhone looks like



Just saying.

J.

[LesterBoffo]

If you really want to get 'into' the actual sim of being a spectator, there's Detlef Faber's nifty Walker figure. Which is now in it's own 'aircraft class' so you can load the walker without using an aircraft, and just wander around the field, watching stuff.