b393ce0a93
Players can use sneak and/or aux1 for far-zooming, so they can far-zoom while treading water in an ocean to look at distant islands, instead of using the aux1=climb down hack, which causes all sorts of weird issues with other features that try to detect movement controls
66 lines
2.8 KiB
Plaintext
66 lines
2.8 KiB
Plaintext
========================================================================
|
|
HOPELESS: Problems that look like they might be unfixable
|
|
------------------------------------------------------------------------
|
|
|
|
#### ##### #### # # ###### ##### ####
|
|
# # # # # # # # # # #
|
|
#### # # # # # # ##### # # ####
|
|
# ##### # # # # # ##### #
|
|
# # # # # # # # # # # #
|
|
#### # #### # ###### ###### # # ####
|
|
|
|
These are things that I would like to have fixed, but realistically
|
|
probably can't do anything about. SOME may also be specific engine
|
|
issues referenced in WISHLIST, but others may be so fundamentally
|
|
broken or unrealistic that they may not be worth pushing as an engine
|
|
issue.
|
|
|
|
- Variable step size consequences
|
|
- Instability of entity physics at large scale, i.e. cannot
|
|
aim a catapult
|
|
- Optomechanic timing issues; I'm unwilling to sacrifice optic
|
|
timing lockstep, but I can't enroll every other system in them.
|
|
- Doors are a possibility but would require refactoring
|
|
out a "virtual node transaction" sort of object that
|
|
could be hella complex.
|
|
|
|
- ABM performance sucks
|
|
- The scheduler distributes work across time very poorly
|
|
- We really need the ability to control scheduling from lua
|
|
- C++ is too inflexible
|
|
- We need to be able to process in batches instead of only
|
|
per-node dispatch
|
|
- neighbor check really could be made more specific,
|
|
e.g. a mask to select which neighbors
|
|
|
|
- Liquid issues
|
|
- Z-fighting against normal nodes; no way to both show inside
|
|
surface against air AND avoid z-fighting.
|
|
- Multiple kinds of water don't "mix"
|
|
- Because only nodes can produce light, dynamic lights require
|
|
use of fake air-like nodes, and because liquids don't mix,
|
|
I can't make fake water-like nodes for underwater dynamic
|
|
lighting, thus all dynamic lights do NOT work at all
|
|
underwater.
|
|
|
|
- Can't do the sky navigation thing fully
|
|
- Client resource GC for computed textures could fix the memory
|
|
leak issues
|
|
- Multi-layer skybox with "parallax" layers would allow me to
|
|
move things around without building new textures (would be
|
|
faster too)
|
|
|
|
- Entity rendering performance is painful
|
|
- visinvs have an outsized impact on rendering perf, making
|
|
large storage areas painfully slow
|
|
- would like to be able to make more nodes ents, like doors,
|
|
for animations
|
|
|
|
- LBM and node timer "leaks" can happen, necessitating a lot of ABMs
|
|
- If I could reliably discover the existence of nodes and
|
|
ensure some process is pending for them, then I wouldn't
|
|
need these.
|
|
|
|
........................................................................
|
|
========================================================================
|