The top one favours visual simplicity, while the bottom one more accurately represents how ramps function in game in terms of fluid mechanics. They both work and make sense. It's just a matter of choices now.
(
Nearly caught up on this thread after a long lag. Plenty of room to have been Ninjaed, though.)
In general I agree with the need for fluid movement to be accommodated (1-3/7
th of a liquid in a full-depth dip should not be seen wetting up and over a half-height ridge in a one-tile ramped-channel between that sunken area and any adjacent area it could spread into) but if the double-steepness ramps needed to drag the groove down to the same floor-level as the sideways-adjacent are as traversable as the standard type (for walking or wagoneering passage) then it might lead one to believe that it is practicle to connect a 2z difference of height with two double-steep ramps (one reaching down to the tile-centre, the other reaching up to there from an opposite
or adjacent side, or something more complicated involving the corners).
I would also be happier with the flatter version of isolated ramp-spires (flat on all sides, centre at +½Z) and ramp-holes (..., centre at -½Z), as they do not functionally impede movement even as much as the implied lump in the ground or pothole, though the creature that routes over such a pothole (or into, to perhaps dig down/drift-mine further from that spot)
does switch Zs. (Not so for a ramp.) Ditto, respectively, with ramp-trench/ridge across a path, including with the extraordinary behaviour of a wagon (extending into the trench wall, with some subsequent effects that produces henceforth suffered or exploited). Half-depth(/height) movement of the ground down(/up) in one tile when moving between two same-height floors that connect through (in all kinds of combinations[1]) would seem to work better for that.
And not all sides of all ramps
must smoothly connect to sides. If you set a miner to stand in a ramp-pothole (centre at -½Z) and channel straight down another level then you have a 'ramp' 2Z down, but it could be as much centre @-2Z with edges/corners all at -1½Z rather than -1½ and -1Z respectively (original pothole with -1Z shift). If you go to the edge of a cliff-edged plateau (or down the wall of a cavern) digging a 'wellshaft' down the edge to send a new miner down (not trivially reversibly) to the base of the rock wall you
want the open edge(s) or corner(s) that arrive at the base level to markedly align with (or slope) that bottom, and the rest are irrelevent.
Consider various other junction plans (mid-rock, but could be at least partially open-air in rough outside ground)...
▓▓▓ ▓▓▓ █▓▓ ▓█▓ ▓▓▓ ▓█▓ █▓█
█▼█ █▼▓ ▓▼▓ ▓▼▓ █▼▓ █▼█ ▓▼▓ Upper Z
▓▓▓ ▓█▓ ▓▓█ █▓█ ██▓ ▓█▓ █▓█
▓█▓ ▓█▓ █▓▓ █▓█ ▓▓▓ █▓█ ▓█▓
▓▲▓ ▓▲█ ▓▲▓ ▓▲▓ ▓▲█ ▓▲▓ █▲█ Lower Z
▓█▓ ▓▓▓ ▓█▓ ▓█▓ ▓██ █▓█ ▓█▓
(█ intended as passably dug/open, ▓ as unpassably undug/wall, but I think you could switch the two and still have at least some valid passages.)
The ramps you need lead between level, without even the complications of adjacent ramps, can be crazy or otherwise challenging (especially the +/x transitionals to the right).
IIRC, you can also 'lean' a ramp up a level against something with dynamic blocking behaviour, as well as fixed wall. So something like a floodgate would allow ramp-upping when closed but make it a nullified ramp otherwise. (If not switchable, building/removing a wall on a given spot like that could change the visible and practical ramp profile.)
And that's before you start to think about letting the below-half-depth water flow 'naturally'.
If I was starting from scratch on this problem I'd suggest
not allowing one-tile ramp-gaps (iteratively propogate the digging-out, manual or in initial landscaping proc-gen, so that it is two or more), or else make the liquids respect the 3½ liquid levels of dam-depth (4+ depth flows over with 1..3 effective depth over the ½Z channel point(s), lower than that and it doesn't flow onto that ground at all).
Or fudge it. Pretend it seeps through anyway. Or assume a 'gutter' (too narrow to bother feet/wagonwheels) in the central dip.
Ditto, water behaviour across a 'ramp ridge' (when constructed, you can have a whole area left with ramps, where digging out from native full rock/soil/sand would 'evaporate' all but a select few lone spires, depending on dig-order) acts as if there's no obstruction so instead of a broad 'pillow' of raised ground it would be more like a small herd of terrapins hunkered down. Water would flow between their shells (ostensibly, if not yet graphically).
Like everyone else, I've thought alot about this. Even did some rendering (a bit in Blender, most in POV-Ray) where I needed to decide how to handle some of the odder edge-conditions. Given that, all I can be definite about is that there is no easy solution that works entirely well for me. Which is why I'm probably happy to accept any half-way decent solution that someone else makes into some kind of solid reality. Won't mean I'm not going to grumble at all, of course.
[1] I don't envy Mephday's problems with this. There are far too many combinations of each edge(-centre?) or corner-point possibly needing to be tied-high, tied-low or else allowed to float wherever seems best - at the very least considerations about not even being passable on that vertex, a choice of which adjacent level to tie passage to or variously co-dependent with adjacent ramp(s).