Player Features and Ideas Discussion

How does it feel to have your space repossessed by your masters?

Maybe this way:

Everyone start drone assist 0.

After learning small drone assist Level 5 (1x), get 5 assist.

After learning medium drone assist L5 (4x), get 5+25 assist.

After learning big drone assist L5 (8x), get 5+25+50 assist.

can be more ?

Meh, doesn't change anything for the majority of the player base, who cares.

it's not even that using more drone bunnies is an inconvenience. it's just that the original post makes no sense. he made special mention of not wanting to affect incursion runners, yet clearly contradicts that with a 50 drone limit.

Specifically, he says that he wants to preserve incursion drone managers. Even with this limit, they're preserved. Even with half the limit, they'd have been preserved. You still have one guy controlling an entire cloud of drones.

you don't because there's still 3/4 of the drones unassigned because he's at the limit of 50 drones...

…and that second cloud of drones is controlled by a second guy. The drone managers are still preserved. No amount of narrowmindedness can remove this simple fact.

So yes, yes you do.

Yeah, though it always depends how bad bad means. Python clearly has no, let's call it state-of the art multi threading support. However not all problems require that. If you happen to have a specific problem that needs state-of the art multi threading, then Python is a bad fit. Not all problems are such though, so Python is a perfectly fine language for solving a huge amount of problems out there. If you require some multi-threading, then there are ways around that where you can get satisfactory results with minor adaptations. Your link (which is excellent, thanks!) shows a few options:



While the link is great (and so is his preceeding entry linked in the head of that post) I find that the conclusion of reading the article however is basically that the GIL is still not going away and that is still very unlikely to change any time soon. It also gives an overview in such sufficient detail that it's clear that none of those solutions will help EVE. I have no idea how much data a big fleet fight is (I mean the in memory state representation of a fleet battle) but I just assume it's several houndred MB worth all in all (including Python management overhead). No solution can really deal with that. multiprocessing is fine with splitting of a process that does its thing and has limited contact with its origin parent and brings data back. What you need in a big single battle where probably 99.99% of processing happens in one single grid is calculate and iterate over all actors repeatedly, changing state based on proximity, damage, etc. in short: work massively parallel on a huge memory region without any corruption. Single threadedness guarantees that, but does not perform. Multithreaded you need the support of the language for proper isolation while still ensuring stuff gets executed in parallel. And there is nothing in Python for that.

So while there are solutions for other problems, for this particular one we want here, there's none given. Worse, it paints a bit of a bleak outlook so you're on the safe side betting there won't be a solution for it anytime soon.



Which is basically why I pretty much expect a full rewrite is required and thus I am saying that this won't get fixed ever. Even a minor break-out of a single functionality to offload into a seperate process (brain in a box) is long overdue. CCP seems not to care enough to assign serious resources to such undertakings - they are not really visible and marketing sexy. I mean they just had another great news event so apparently it works well enough in the state it is in. Additionally, the subject matter at hand is too technical and complex so to see that nothing will help except true paralellism in the server processing for anyone who did not do actual work in a field where you get bound on single thread execution speed on current hardware requires very specific in-depth knowledge which simply is not widespread. So they get away with vague responses that they work hard at solving it and "looking into it" which placates the majority asking for improvement in this area.

It won't come though. Not tomorrow, not in a year. Not in 5 years. Unless magically someone fixes the GIL problem in Python which the entire Python community tries to solve for.. what 14 years now? Or like I said, they rewrite the entire server code for fleet combat on grid. Potentially from scratch. Not. gonna. happen.

…which he is.
Thus, Rise's comment about wanting to preserve drone managers is entirely correct.

You can twist it as much as you like, but it's still the case.

Exactly why I made my original response. The change does affect incursions in a negative way (while devs stated they don't want that) and it needs to be pointed out.