Networking protocol version 3

[Minoru]

Hi there!

Turns out I didn’t fit all the necessary features in the protocol version 2, so we need yet another iteration. We didn't document our discussions, and I'm not really sure my memory is reliable enough, so it might so happen that I didn't remember all the details. To prevent situations like "we agreed on that but you didn't incorporate it in the protocol", please review the list and comment if anything is missing:

1. Relative positions instead of absolute ones. This one is bigger than you think. Other than the obvious changes to the move and there you see messages, there would be changes to:

   • who is there? message: x and y would be relative, too. I wonder if we should just "fix" the "radar" at the robot, i.e. do not specify any coordinates when we "look around"?

   • bump would return your new position relating to the old one.

     This change also brings a problem: how would agent know if he's at the border of the map? Should we add new message, or is bump sufficient?

2. Implicit turn on move. Generally, it's what hub should take care of, but I think it's important to document in the protocol as well.

Please answer the questions I asked, state your opinion on the changes I propose, and add details I've missed (Vlad, I'm looking at you)!

[Minoru]

Issue #15 describes some "error message" without any details. Should it be incorporated into protocol version 3?

There's a problem with turn message: its parameter is absolute value (orientation specified in degrees). Should it become relative, too?

[Minoru]

Draft version of the new version of the protocol is available. Please, please review it and give me some feedback.

[Minoru]

Some of most the notable and debatable decisions I made and the problems it may cause:

• move accepts only one parameter, namely the distance. Agent moves in the direction he is facing at the moment. That means that we can't have robots that look sideways while moving forward.
• who is there? only asks about the objects around the agent, i.e. we can't look at the other side of the map. I wonder if we should remove 'radius' parameter as well — we have visibility radius in the robot settings already, don't we?
• instead of x and y coordinates we now use azimuth (relative to the agent's orientation) and distance. The former is specified as an integral amount of seconds (1/360th of the degree), the latter is an integral amount of pixels. That may pose a problem: first of all, we can differentiate no more than 360 * 360 = 129 600 points on the arc, which limits us to the circles no more than 129600/(2*pi) ≈ 20626 pixels in radius. Second, we may experience wrong (t, r) → (x, y) transitions due to the floating point rounding errors.

If any of those decisions are an issue, please let me know.