I encourage you to read the whole thing, but the standout quote for me is near the end:
> One thing I learned from discovering The Elm Architecture is that it is really lovely to be able to show up in any codebase and know what is going on. I think custom operators detract from that enough that they are not worth it for the whole ecosystem, even if they are great for specific individuals.
Having talked to people at work who had to remove custom operators when they upgraded from 0.18 to 0.19, they initially didn't love dropping them, but getting rid of the syntax cliff for new hires at the expense of terseness for the old hands was a decent tradeoff. Personally, I find it can be easy in an ML family language to get a bit wrapped around the axle trying to write the most terse, pointfree implementation of something.
> If they try to register their lien once the owner of the vehicle has changed, it can not be accepted.
This is true in practice with houses, not in theory. Here's an example (differs by jurisdiction) of how the deadlines work for filing a lien: https://www.levelset.com/payment-help/question/can-we-lien-a.... The title search should include "are the sellers the defendants in a lawsuit wherein if they lose, the judgment results in a lien on the property." We actually went under contract on a house where the owner's name matched someone in bankruptcy and triggered this. We knew the guys were different (ages, addresses) but our mortgage lender needed the title company to figure out their mistake first. So the lien attaching to a house that sells in the meantime is a really extraordinary circumstance: closings usually take 30 days so what are the odds that something gets filed in that meantime without already being underway? If the lien does attach to the house, then the buyer would have a tort against the seller for putting them on the hook, but the mechanic would have a backstop for getting paid in the form of the lien. The buyer could get a judgment against the seller. But that'll all be handled by the title insurance if it ever comes to that.
> The person who takes the debt is responsible for the debt. If they sell the car then they pay the debt with the money from the sale, if the sale doesn't cover the debt then they still owe the rest.
This is how it really works in practice. The mortgage company won't let you buy the house with the liens still attached. The debts will get paid out of the purchase price of the house. Who gets paid what out of the sale price is already determined before the actual closing date, and the mortgage company writes separate checks to everyone who needs to get paid (I think). The insurance is there in case something very out of the ordinary happens. Even the cost of paying someone to figure out who gets what is part of that insurance policy (again, I think).
> If the owner fails to resolve their debts when they sell that should be their problem not the buyer's.
It effectively is. The title search before the closing reveals all of this and pauses the sales process. Nothing proceeds until they come up with a plan to pay it off. The insurance is only for extraordinary circumstances where someone messed up the recording of the lien.
> There needs to be a system for registering and searching liens, it needs to be publicly accessible and the lien needs to be registered before it's valid.
This would be the best. I'd love it if it were federal so you just had to search one thing. Bankruptcy is federal, so why not this? Even better: if the system could record the house going under contract. You're a roofer who's been lazy suing somebody. Every Monday you upload the CSV of addresses that owe you money. One of those days, you see "this house is now under contract and you have 5 days until the window to register lawsuits against the sellers closes" You call your lawyer and tell them to start the lawsuit for the unpaid debt. Sellers and buyers get notified that the title isn't clear and there's a suit worth $X to work out.
This then produces the problem of people monitoring every address in their town and threatening the parties to the sale with a baseless lawsuit in order to pause the sale. You may have to bring back corporal punishment to stop such individuals. Hmm. Much to consider.
> the property can only be sold if everything is OK with it (no litigation, liens, etc), and taxes are owed by persons? Is it different over there?
This could vary by jurisdiction, but as I understand it, taxes and liens are attached to the property itself. "Clean title" can be a contingency of offer: buyer can back out and get back their earnest money (aka deposit) if the property has liens/encumbrances that are not written down in the sales contract (example clause at the link at the end). When you buy the place, you get title insurance, often mandated by your mortgage lender. The title insurance company does a title search on the property to find liens and owed money on the property and then sells you an insurance policy saying that they'll make it right if they missed anything during their search. This is because your mortgage lender never wants to be second in line to get their hands on the property to recoup in case you default on your mortgage. Liens on the property should be easy to find because they're supposed to be registered with the local municipality: maybe the city you're in, maybe the county, maybe the state, idk I think it depends. In practice, maybe some roofer/plumber/landscaper forgot to do that and now you have a problem you didn't know you had. That's what the insurance is for. The property _status_ is not knowable so much as the _status transitions_ are knowable: when was a lien attached or removed from the property, so that's why it involves a private company looking it up. You'd think it'd be a public good, but it's not. Odd.
As an example: when I bought my current place, the previous owner was financing the furnace which included free annual service from the installer. He wanted us to take over the payments. We asked him to convey without encumbrances, meaning pay off the balance with the furnace company before we'd close on the house. If he had refused, we could have backed out of the sale because our offer said that we were only willing to buy without owing anyone anything.
If you'd like an incomplete list of everything that's downstream from Elm, I made one the last time a TEA library got to the front page here: https://news.ycombinator.com/item?id=47678424 The creator of Derw shows up one comment later to remind me of one that I was forgetting.
Slightly important point: it's still safe FFI! Everything that crosses the JS/Elm boundary has to be encoded/decoded from serialized JS Objects. So it's hacky BUT it's not actually unsafe, which is kind of funny. I have an example where I use it to effect locale-aware sort. This repo is a benchmark test of that: https://github.com/perkee/elm-js-sort-ffi-speed-test/tree/ma...
Haha we may have worked together. Equatable has been a theoretically huge, actually OK hole in my opinion because it _can_ lead to runtime crashes but never has in any code I've written (\_ -> () == \_ -> () is a runtime crash). So I'm glad to see something more robust than "just don't do that." And "hashable" as the real thing you need to implement Set/Dict is a good idea imo. I'm sympathetic to keeping "comparable" closed and even smaller than it is for the reasons in this discussion [0]. In my perfect world, String would not be comparable as-is because string comparison should be locale dependent. Right now string comparison just ends up using JS's < operator which compares as a list of code units. So I'd like to be able to have a Set of Strings or a Dict with String keys without having the footgun of being able to think that I've alphabetized a list when all I've done is annoy a user by putting Área after Zapato and not between Azul and Barcelona.
But I agree with you that the current solutions for Set/Dict all have one problem or another: You can use elm-sorter-experiment [1] but then you are writing a sort function and passing it around and maybe you don't want the overhead of that. You can switch to the Lamdera compiler and use containers [2] but then it's a different compiler and a you need to tweak your options slightly to compile an Elm project as an Elm project in it and it could get out of sync with the Elm compiler (extremely unlikely though). You can use any of the list/dict implementations that are really just lists with O(n) everything under the hood. So maybe this hashable type is going to make things better, idk.
there is also web components (best way to do a "copy this to clipboard" button in my experience) and secret back door synchronous FFI through overloading the Object prototype. In all cases the communication is gated through decoders/encoders that can fail; you can only pass serialized data across the boundary from JS to Elm.
You can also do lunatic things like overload the `Object` prototype and get synchronous FFI by encoding then decoding an object. I do this at work to get locale-aware sorting. I know you know this, but it'll be news to others in the thread.
> a type safe language with a single solid toolchain
Been paying the bills for 5.5 years now writing in one that gets a fair amount of stick for being ~6.7 years with no language version releases (one is planned for this summer though that should make it just under 7). There's one compiler, one standard library, one formatter (with no config options), one linter, and two test runners but one is a rust rewrite of the other one so it's chill. The language being on a timeout has led to all kinds of things like compilers for native (consuming the intermediate representation .o files), fullstack un-forks (no change to syntax of language), fullstack actual forks, really a lot of weird but delightful activity.
It's like me and a few hundred other people, so "unified" is the part that's missing, haha.
it's like reading instructions for using a GUI program (open this dialog, click this checkbox, fill in this value) vs copying and pasting a complex FFMpeg command
If we include the word Latin, then we have speakers of Ladino (also called Judeo-Spanish) and Ladin (natively called ladino). The first is derived from Old Spanish and spoken by Sephardic Jews everywhere but in Spain where they were expelled from. The latter is spoken in Northern Italy, specifically in South Tyrol, Trentino and Belluno.
The names are funny just because you can imagine asking someone speaking a language derived from Vulgar Latin what language they're speaking (or writing), and them answering as a fish would when asked about the water.
You are allowed to try to use the defense in some places, but there's no guarantee that it will work. It is banned in DC and 30 states: California, Illinois, Rhode Island, Connecticut, Hawaii, Maine, Nevada, New York, New Jersey, Washington, D.C., Georgia, Wisconsin, Washington, Pennsylvania, Colorado, Texas, Virginia, Maryland, Oregon, Vermont, Florida, Iowa, New Mexico, Minnesota, Massachusetts, Nebraska, Arkansas, North Carolina, New Hampshire, Delaware, Michigan. Put another way, it's banned for about 76% of the US population. Does it actually work a lot when it is used? Did it ever? Note that the case you're referencing is from 1944, for instance.