For me filmic with zeroed exposure fiddling, no midtone saturation and no preservation looks close to finished (if the shot has no other technical flaws.
I understand this to be what people consider pretty raw or start flat.
Edit @priort its a huge difference because one side of the coin helps beginners and neither side makes any difference to experts.
I dare to state that any current raw-processor is capable of results that are good if not perfect.
If the default pleases anyone is mere luck. Trying to make a tool work for everyone will result in less than mediocre software. Always. Reasonable defaults are a different thing, those are up for debate me thinks. Mind ā¦ āreasonableā ā¦ not āperfectā ā¦ otherwise our cameras would have a single button like a Kodak Brownie and everything would be āpush the button, we do the restā.
All I see is that every raw-processor I ever used had reasonable defaults.
Can they be improved? Sure.
If someone finds themselves constantly changing a certain setting they should put it into their basic preset. Because it has been mentioned ā¦ I disagree with a number of default settings in dt, e.g. filmic (chrominanc), exposure (compensation), denoise (wavelet ftw) and so on and so on. But all of them are personal choices, not a fact.
The amount of research that goes into creating the mass-pleasing defaults in software by large corporations like Adobe, Apple etc is insane. But by creating those defaults they also shape the perception of the public. So either you want that exact look, then buy the product. If you want to express yourself then you have to put in the effort and learn to create your own defaults.
That will be a hard time. If it will be worth your while is up to you.
The key words here are āFor meā and your are right for you that is perfect. I also think its a good choice. But others will have reasons for other choices. This whole discussion shows that there are different expectations and approaches and so in my book this invalidates this notion that the defaults are likely to be anything other that which the developers believe to the best and accurate values of the light recorded and not to push the pixel math into somewhere that can be defined by some physical or rational limit while at the same time giving the user every opportunity to disagree and provide new settings. I think both RT and DT make this possible
Of course. Good enough, reasonable, carefully designed.
If it was expected to be the only option the GUI would turn very minimalistic 
Iām a bit shocked how many times I have to repeat it.
Are they good reasons? Thats a productive discussion. Copping out because subjective isnt how its done and it is done.
There are arguments, workflow, use cases that can help. If someone wants acid trip simulation as default there are arguments againts it.
About as productive as this discussionā¦I have no idea what point you are trying to makeā¦Can you can the word goodā¦its completely irrelevant coming from an individualā¦
Pretty raw is straight out of camera with white balance.
EDIT for meā¦
With no demosaicing?
No color conversion from raw color to a sensible working/output color space?
The answer from @ggbutcher might be yes but I suspect @wpferguson assumed those steps as part of the raw data prep for editā¦
This is a weird discussion, full of sweeping generalisations.
In the old days, transforming a film taken by your camera to a film negative that you could reproduce was a non issue - an elementary process, that rarely was discussed, cos it was automatic. Job done , then the photoshop could start the real work of how best to get this onto paper.
This was true for many photographers, but not all. Some of us used 5x4 film that we processed individually according to the scene contrast. For 35mm film, we took multiple camera bodies loaded with identical film that we processed differently.
In 2021, we are not asking too much to expect that photo processing tools built with open source, provide excellent starting points in image processing.
I agree, but we might disagree on the definition of āexcellent starting pointā. We would all have different definitions, and I donāt think thatās a problem.
If anyone cares, my starting point is demosaiced and colour-balanced, changed to the colorspace of my monitor. Thatās it. No adjustment to contrast or chroma, no sharpening, nothing. Then I decide what adjustments I want. The next steps might be adjusting tones to check for shadow and highlight detail, and desaturate to check overall tonal impact, and massively increase chroma to check what hues the āgraysā really are. Probably none of these will be āfinalā images, but they all provide me with information to decide what the final image will be.
But thatās just me. I can see that other people may prefer to start from images with boosted chroma, sharpened, and so on, aiming for a point closer to a finished āusableā image. And if they can define presets that do this, thatās okay. I donāt have to use those presets.
Why wouldnāt I? For example, when an image has been sharpened, I canāt easily see what points are in focus, and how far out other areas are, so I canāt easily judge which (if any) areas need sharpening or blurring. I accept that this is because I am fussy. Other people may not be bothered about selective sharpen/blur, and merely want an overall sharpened image.
Part of the issue is people mixing up two concepts.
People can argue about whether a neutral or a punchy output is better for number 2, but I think itās clear that a punchy output is generally better for number 1.
Nope just against something that can only be defined or determined by the individual. If its a particular module or function you should propose another value or method and put in a feature requestā¦if it has merit or traction I am sure you will get support.
Are some of you on vacation? 
I come back in a few hours and I see 54 or more new posts. TL;DR after the conversation slows down? 
You have to demosaic to get white balance
Sometimes no working or output color space. Sometimes I turn off the white balance to see what channels are really blown (usually when working on Play Raws). Sometimes I adjust the input color profile for tricky images.
In darktable, maybe. In terms of raw processing, not strictly true.
Hereās what I do:
At this point, this is what Iād consider ālinear RGBā. Now when I view it, itās through the display transform, so thereās that going on. Also, Iāve also usually opened the raw file using the batch-produced proof processing, which includes a filmic tone curve, proof resizing, and a bit of convolution sharpening. But, I usually select the black/white point tool for display for a bit, and consider the linear tone relationship. Iāve had images where the scene was lit such that I could delete the filmic tool and āgo linearā for the rendition.
Typically though, Iāll mostly mess with the tone curve, either tweaking the filmic one, or in the case of reaally high DR images, Iāve been changing that out to a log curve and adding a control-point curve to shape the tone distribution to my whim. If the image has extreme hues, Iāll also switch out the matrix camera profile with my LUT profile made from spectral data.
Since Iāve started using the highlight-weighted matrix metering mode of my Z 6, Iāve had to spend more time with tone curve strategies, because no one curve handles all the scenes. I think this would not be the case if I were exposing to the middle gray, but Iād be using highlight recovery moreā¦
Have a look at this PlayRaw threadā¦
Itās pretty hard to imagine that all the contributors on that thread would prefer the same starting point. What would seem āgoodā to some people would look like a ārip and replaceā to others.
Checking out of this circular argument.
You are the smart oneā¦ 
Once upon a time two people have a chat
A: Letās define a good default ā¢ for raw processors.
B: But what is a good default ā¢ ?
A: That is exactly what we are going to discuss.
B: But there is no good default. Good is in the eye of the beholder.
A: Iām sure there is. Camera manufacturers and commercial photo editors spent years of research to produce this good default ā¢. So there is indeed a good default.
B: Yet they produced many different defaults. Some people like Fuji look, some like Nikon color, etc. So there is no a single trully grand good default ā¢.
A: The point is not producing a greatest look, but an acceptable starting pointā¢ for further editing.
B: Okay, so we are not discussing a grand or good default ā¢ anymore. Great. So what is an acceptable starting point?
A: That is exactly what we need to define here.
B: Well many people have difference preference on their starting point based on their workflow. Some prefer more āunprocessed lookā to not anchor their creativity to something pre-baked. Some prefer a half processed look so they get more productive in the area they want to retouch. But we have many tools for both approaches, donāt we? Even the FOSS raw processor which starts with more āflat/unprocessedā look allows you to create preset for your own default.
A: But if we just start with something acceptable, it shouldnāt harm anyone. The beginners will produce a nice result immediately, and the experts can still do what they want.
B: Sorry what you mean by āsomething acceptableā? You mean āhalf processedā look?
A: Ummā¦ yes.
B: But I donāt like that starting point.
A: But I like it. I think most people will like it too. If you donāt like it, you can always disable that feature.
B: Why would I write a feature that Iām going to disable?
ā¦
Legends says that this conversation is still going till now in 2021.
A lot of things can be done. But the point is, that many users and developers do not want this kind of change. The others can change their workflow to Filmulator plus a second software, make their own presets or fork a software. But an āendlessā discussion that tells others, my opinion is right and your opinion is wrong, is not in the least constructivce.