Re: [Nel] TCP vs. UDP

[Vincent Archer]

According to Zane:
> I don't believe that this is entirely true.  Let's look at what we need out
> of a protocol for NeL:
> 
> 1)  Know what data HAS to be delivered and what data doesn't really.

That's priority, and bandwidth management. Not really TCP-vs-UDP relevant.

> 2)  Know what data can be overwritten by new data if we recieve a large
> chunk of data

The latter requires you to look at the semantics of the data, not just the
syntax. You have to posit specific types of data, and, for example, you
don't have to retransmit data type X about object A if it has been already
sent out, but you have to retransmit data type Y about the same object A.

This quickly require you to tie the protocol to the exact game being made.
A current 3D coordinate/movement vector does not require retransmission
(or, at least, not the retransmission of the same data; you may send back
 a more up-to-date version), but it automatically implies your game has
moving 3D objects.

The example above is simple and obvious, but you quickly have to tie a lot
of knowledge about the game being made *into* the protocol. Making the
protocol generic enough to handle an "unspecified" game is *hard*. Hard
enough that I contend that using TCP in that case might be a good
compromise.

Of course, the more specific (to a game design) your library is, the easier
it is to make it UDP-optimised.

-- 
        Vincent Archer                  Email:  address@hidden

All men are mortal.  Socrates was mortal.  Therefore, all men are Socrates.
                                                        (Woody Allen)