[tptacek]

This is a good, well-written explanation of the most reasonable, mainstream aspects of Urbit, leaving the impression that it's essentially like a P2P version of AWS Lambda.

But of course, that's not all it is.

It's also a ground-up reinvention of mainstream functional programming languages like Lisp, built on the foundation of an abstract virtual machine in which the "decrement" operation can be performed natively only by incrementing a number until it's 1 less than the current number. All of these concepts have their own weird names; for instance, the constant-time version of "decrement" (this is a programming environment that goes out of its way to achieve constant-time decrement†) is an example of a "jet", where a jet is apparently a non-native implementation of an algorithm that can be expressed but not efficiently on the VM that they've chosen to build their entire system on and it just gets weirder and weirder from there.

You don't have to memorize the new names they've come up with for most of the ASCII punctuation characters, like "gal", "gar", and "hax" for "<", ">", and "#", "but it helps". A normal engineer's reaction to a system that tells it that it will help to remember a new name for the pound sign is to ask "Why? What fresh horror lurks in the deeper meaning of your new name for the percent sign? And why won't you tell me before I commit to this system?"

Stuff like that would be bad enough, but the founders ideological views are also infused into the system, and those views are not mainstream distributed systems engineering views. For instance: the most available first-class address in the system is in a 32 bit address space. Why? Not for efficiency, but because the authors believe there aren't and never will be 4 billion human beings on the planet worthy of having a first-class address in their system. This, by the way, is their actual response to the objection of an overlay network with 32 bit addressing.

I agree that the best way to explain this system while encouraging people to engage with it is to distill it down to anodyne concepts and then sprinkle "decentralization" on top. But of course, this system isn't really that. From the proteins of its cell membranes to the DNA in its nucleus, this is a system that coercively projects the idiosyncrasies of its founders into everything it comes in contact with. It's decentralized and free in the sort of way its founders would explain with a 3000 word paper that invented 50 new terms in its abstract.

People confused about Urbit are probably not confused about the value of serverless computing or of overlay networks. Those are pretty straightforward concepts we can all get our heads around. There is something deeper that is challenging about Urbit.

† Rather than, you know, just deciding to have it.

[chc4]

I'll admit, Hoon is weird. But complaining about Nock is like complaining about the JavaScript VM. You can JIT it and replace common functions to replace with calls to C functions instead. Nock defines the /results/, not how to get them. Jets are basically just because the Nock spec didn't want to have to outline a bunch of primitives that don't really matter, since they would be replaced with native calls anyways.

As for ideological views: there are 4 million planets. They are, essentially, houses, but each one of those planets can also issue 2^16 "moons", which can themselves be their own people. I don't know the reasoning behind why 2^32, but I'm also not going to ascribe the reasoning to human worth any more than I would the author of IPv4. There are also 2^64 comets, which are their own ships outside the trust heirarchy, which really just means that they have a default karma of 0 because they have no cost, and so can be Sybil attacked - not that they can never get a positive karma, or that you couldn't use one indefinitely instead of a planet (just that it's be unwieldy). Saying Urbit can't scale bigger than IPv4 is kinda like worry that Hoon can't be programmed in Chinese - there are probably bigger problems, and it could be solved by just doubling the amount of galaxies. Hell, Urbit is open source. There would be push back against doing it, but there is nothing stopping that from being patched in tomorrow.

The most common thing I've see n called fascist in Urbit is it's identity system, and this Urbit+Ethereum spec seems to be a direct solution to some of it. Now, even buying a star you don't have to trust the owner of the galaxy you bought it from, because it's a redeemable token.

[tptacek]

Whoah, hold on. I hate to zoom in and nitpick, but I'm not guessing when I provide that reason for the 2^32 bit address space. It's the official stated answer from the project itself.

And I'm not saying the problem with that design decision is that it's amoral (let's stipulate that it isn't). I'm saying that it's extremely unconventional, a landmine of orthogonal ideology buried in the architecture of the system. Before committing to building a new product in someone else's framework, I'd want to know about all such weird decisions, and about why they are that way. Maybe Urbit's principles are compatible with the way you think about every possible aspect of how your system will interact with any of its users ever (since that's what you're buying into when you literally adopt an entirely new network to build against and deploy on). But maybe not!

Most framework designs coercively project opinions about whether the efficiency of register-sized addresses are worth the flexibility tradeoff, or whether it should be easier to define new URL patterns versus making it easier to dispatch the full complement of HTTP verbs, or whether closures are first-class or sum types are worth the complexity. Not a lot of programming environments ask us to determine what double-digit percentage of the world's population is too irresponsible to deserve an address. Again: their words (almost; I changed the order).

I'm trying to be careful not to use loaded terms like the f-word you just used, by the way. What I think about Urbit's founder is not necessarily the same as what conclusions I'd draw about this particular system, which is, I think, fatally flawed on its own demerits.

[chc4]

I haven't seen that as the reason behind it (something about 4 billion planets being a good limit, maybe), and assumed it was hyperbole.

2^32 "houses", even if it was ideologically motivated, is a design decision that makes sense at first glance, is easily worked around (and was designed that way), and can be trivially changed. Some of Curtis' politics were visible in the Urbit first draft (naming galaxies after ancient rulers being the most egregious...), and both weren't important and have since been changed.

Thinking about it, 2^32 planets is one of the only weird design decisions that /would/ be ideologically motivated, other than the general dislike of Haskell or UNIX that birthed it. All of Hoon and the kernel are basically built from Nock first principles, which is both insane to read and amazing, and I'm not sure that you can say Curtis' ideology is the DNA of them. Urbit has tons of weird things in it that could make you drop it (like all of Hoon, even though most of it does make sense and is easy to learn, I promise! I'm not crazy!), but I don't think Curtis is the fatal flaw.

[ineptech]

> I'm not guessing when I provide that reason for the 2^32 bit address space. It's the official stated answer from the project itself.

Link? My understanding was that the official answer was "there should be about as many urbit addresses as adult humans so that they're too expensive to spam from, and if urbit is ever so wildly successful that we start running out, it won't be that hard to add more." (though I don't remember where I picked that up either)

[tptacek]

A 32-bit planet is a tool, not a toy. Like a car, it's a device for a responsible and independent adult. There aren't 4 billion cars in the world, nor 4 billion independent adults.

If you aren't an independent adult, and you don't need or even shouldn't have unconditional digital freedom (no one's 8-year-old daughter needs unconditional digital freedom), a moon from someone else's planet is fine. (Even most of today's independent adults don't complain enough about being Facebook's moons.)

It's literally a bullet in their "objections" post.

[ineptech]

You said the founder believes "there aren't and never will be" 4B people "worthy" of an urbit planet. I'm asking where you got that "never will be" part.

I think your reading of that bit you quoted is extremely uncharitable. I don't see where urbit is passing pronouncements on anyone's "worthiness". The way I read that quote is, "an urbit planet is server software, and there are way less than 4B people with a server to run it on."

[pcmonk]

There are about 5 billion adults[0], and they don't all need separate planets. Married couples, for example, may not.

That's beside the point, though. Using scarcity to create value is not "extremely unconventional"; it's about the most ho-hum idea in Urbit. If you want a 128-bit address, you can use those in Urbit just fine. The only difference is that they don't have a default-positive reputation (where "reputation" is an abstract concept -- in practice, e.g. some channels wouldn't let 128-bit address post comments)

[0] https://www.census.gov/population/international/data/idb/wor...

[njarboe]

4,294,967,296 is the number of planets which is a much more than 4 million. Think about a planet as a piece of land more than a phone number. How many people on earth own a piece of land? And when will that reach 4 billion?

Also, if Urbit becomes popular other groups will start similar systems and they will be able to interact.

[grzm]

I believe the 4 million in your parent's comment is a typo. Further along in the same comment they refer to 4^32 and down thread they refer to 4 billion.

[yebyen]

Like IPv4, it's a theoretical maximum that will be hard to reach because the number of allocations are divided up into blocks... Universe(Galaxy / Star / Planet / Moon) are the top 64 bits of the address space, so there is a 2^32 space for planets, and the top 8 bits of each segment followed by all zeros would refer to a (Galaxy / Star) of which each are also addressable, single machines.

So yeah, 2^32 planets (minus 65536 stars (minus 256 galaxies)) and each planet has a plot of "land" that is 2^32 wide to allocate moons out of. What happens if we run out of planets? Then people will start living on moons, simple. Or comets (the remaining 64 bits in the 128-bit address space are for comets, which are not comparable to land or homesteads.)

[grzm]

I'm not sure I understand the relevance of your comment. This is the motivation for mine:

tptacek> [T]he founders ideological views are also infused into the system.... [T]he most available first-class address in the system is in a 32 bit address space.

chc4> As for ideological views: there are 4 million planets. They are, essentially, houses, but each one of those planets can also issue 2^16 "moons", which can themselves be their own people. I don't know the reasoning behind why 2^32....

njarboe> 4,294,967,296 is the number of planets which is a much more than 4 million.

I read 'njarboe's comment as a correction of chc4's use of "4 million". 4,294,967,296 is 2^32 (rounded to 4 billion), which 'chc4 mentions here as well saying 4 billion another comment.

If I've missed something, please do correct me. I just think 'chc4 meant to type "4 billion" given the their other correct use of 2^32 in the same comment and 4 billion elsewhere (https://news.ycombinator.com/item?id=15300641), and 'njarboe's misinterpreted 'chc4 as misunderstanding, rather than fat-fingering. I'm not saying anything about the choice of 2^32. If you read that as otherwise, how can I improve my writing to make that clearer? Was there a particular phrase that you found easy to misinterpret? Something that made you think I was saying anything more than that? Do you think 'chc4 meant "4 million" and not "4 billion"? Are there actually only effectively 4 million (not billion) planets? As for my interest in continuing this thread, it's not that care about the discussion topic, but rather whether I'm communicating (reading and writing) effectively.

[yebyen]

No, it was clear, the point I was contesting was the first point that you quoted. I could have been clearer on why I was commenting.

> The most available first-class address in the system is in a 32 bit address space.

The most available address is a comet. What is the meaning of a "first-class" address? They are all addresses, the motivation of the founders and the concept of a first-class Urbit address is what I take issue with. There is currently very little that you can't do with one of the 0::(64bit) addressed comets, except for "spawning child addresses." The other differences are all theoretical solutions to problems (spam) that we Urbit users in great part just don't have yet, so the idea of a first-class address is basically theoretical too. Having a planet is more like having a /24 if you want to draw parallels to IPv4, but... not really a valid comparison.

Comets are the most available addresses. There are 2^64 of them and it only costs a few CPU cycles to generate you a private key. They have no subordinate "child" addresses. Otherwise they are first-class citizens and you can have as many of them as you want. Go ahead and take your share of that pool of 1.8446744e+19 addresses.

I don't know who would pay $10 for a planet. I know they don't want you to get more than one, in case they need to impose spam controls on these "lower 64" free addresses. I know lots of people have paid for stars, but I don't personally know anyone who didn't get their planet for free at this time.

(Disclosure: I own a galaxy, I don't sell planets, and I did not pay any amount of money for my galaxy. I was in the right place at the right time, and I helped test early releases. I am skeptical that it has the same value as the galaxies that are still held by the founders (it is priceless to me,) but they are active contributors and there is a good reason to give them money if you want to see Urbit development continue.)

There is no such value to be extracted by paying ~del's discount planet shack for stars or planets. I am hardly even a casual hoon programmer.

[adiabatty]

> the most available first-class address in the system is in a 32 bit address space. Why? Not for efficiency, but because the authors believe there aren't and never will be 4 billion human beings on the planet worthy of having a first-class address in their system. This, by the way, is their actual response to the objection of an overlay network with 32 bit addressing.

One of the things they want is for identities to cost some nontrivial amount of money, about $10 or so. This mostly solves the spam problem because it's unlikely that spamming will get you $10 worth of value before you're blacklisted by your parent star and it refuses to route your packets.

Can you think of an identity count that would preserve moderate scarcity like this? Or would you have the address space be indefinitely mineable, kind of like bitcoin?

[ubernostrum]

If it hasn't already, the total number of humans with mobile phone service will exceed 2^32 within the next year or two, and phone service typically isn't given away free out of the goodness of the phone companies' hearts.

So the mobile industry is already past or very soon to be past the 32-bit space with paying customers.

[adiabatty]

Yeah, 2^32 was _clearly_ chosen based on what makes for nice, round numbers in binary, not based on reasonable population and cell-phone-use projections.

On the other hand, Urbit might never become popular enough for this sort of thing to really matter. We're at about 7 billion people now and the second derivative of the world population is negative, so it should level off at some point. If only 5% of 20 billion people care enough about this sort of thing to actually run a planet, then that'll be only 1 billion people, or a quarter of Urbit's 4-billion address space. 5% doesn't seem unreasonably low to me, either; app.net probably didn't even get 0.5% of Twitter's user count.

[ubernostrum]

Urbit might never become popular enough for this sort of thing to really matter.

And decisions like this might be a part of why it never becomes popular. Elsewhere in the thread people are suggesting that, say, children don't need their own "planet" or whatever it's called, and that they could be one per family and still service the whole world. But that bakes in an awful lot of assumptions about culturally-specific ideas of family and stability of family units. Granted, I think those are culturally-specific ideas Moldbug would love to impose on the world if he could, but this is why people keep saying that even the technical decisions in Urbit are pushing specific political positions of his.

Also, at this point anyone who deliberately designs an address space to be too small to handle already-existing popular things is probably not making sound long-term technical decisions to begin with.

[Nition]

Have they given a reason as to why it's not larger than 32-bit?

[yebyen]

The address space is 128 bit. You don't need a planet, but for a theoretical $10 price tag, why wouldn't you get one?

(FWIW I have never seen anyone pay $10 for their planet.)

[ubernostrum]

Yes, and it's been quoted and referred to multiple times in this discussion thread.

[adiabatty]

> You don't have to memorize the new names they've come up with for most of the ASCII punctuation characters, like "gal", "gar", and "hax" for "<", ">", and "#", "but it helps". A normal engineer's reaction to a system that tells it that it will help to remember a new name for the pound sign is to ask "Why? What fresh horror lurks in the deeper meaning of your new name for the percent sign? And why won't you tell me before I commit to this system?"

Odd, I saw the rationale early on, when I was just reading the documentation for Urbit. If you've ever seen Hoon code, you'll notice it has a lot of punctuation in it. Have a look at tumblr.hoon[0]. If you want to speak with fellow Hoon programmers in person and communicate fluidly, it's nice to have a compact, standardized-within-Urbit way of saying things like "$%"[1] out loud.

Now, at this point you're probably wondering "why would they have names that are all punctuation?". As far as I can tell, this keeps the identifiers short and of uniform length. From what I understand, one of the problems that they're trying to solve is that single-assignment code in, say, Lisp tends to flow down and to the right. They're trying to limit rightward visual creep, as I understand it.

Now, I'm not convinced that focusing on standard-library-identifier length is a good idea. Early C linkers could only handle function names that were six characters or shorter, and constraints like the ones in Hoon's standard library sound similarly arbitrary and likely to lead to unintuitive names like "strchr".

[0]: https://github.com/urbit/examples/blob/master/app/tumblr.hoo...

[1]: buccen, which I pronounce as "buck sen". It's used for making tagged unions, which reminds me of Erlang/Elixir {:ok, Value}/{:error, Reason} value-passing style.

[occultist_throw]

> You don't have to memorize the new names they've come up with for most of the ASCII punctuation characters, like "gal", "gar", and "hax" for "<", ">", and "#", "but it helps". A normal engineer's reaction to a system that tells it that it will help to remember a new name for the pound sign is to ask "Why? What fresh horror lurks in the deeper meaning of your new name for the percent sign? And why won't you tell me before I commit to this system?"

Speaking of that.. I looked again at the reference documentation and the "hello world" (Fizzbuzz). It feels like I'm playing a game of Magic the Gathering on another world, with different physics and vocalizations I cannot do.

https://urbit.org/docs/hoon/reference/

     {$book p/(list {{aura @} moss})}: mold which recognizes a union tagged by head atom.
     {$lamb p/moss q/moss}: mold which normalizes to an example gate.
     {$port p/moss q/seed}: form an iron gate.
     {$gill p/moss q/seed}: form a gill, a wet one-armed core with sample.
     {$gasp p/seed q/seed}: form a gate with burnt sample.

Uhm.. What? I know I'm reading words. But part of it makes me feel like Im also in Game of Thrones. Iron Gate, wet one-armed core, and a burnt sample?

And their demo (Fizzbuzz) is just as illegible. https://urbit.org/docs/hoon/demo/ . I get programming. I know there's concepts in different langs that need understood. But one thing I know is that the demo should be easy to dip your toes in.

And it starts. "Tape" is really a string. "Home Desk" -Im guessing its the urbit home directory for your user? Urbit Pier? No bloody clue. But Im supposed to copy the file (oh wait, is file even used, or maybe its joobpflep?) to it. But at least the "Urbit Dojo" is the shell.

[Fang_]

The reference docs honestly aren't the best place to start when learning Hoon, as backwards as that may sound. I've found the tutorial docs[0] to be much more helpful when first exploring Urbit, and there's a fair amount of good community documentation listed on the docs page[1] as well.

[0]: https://urbit.org/docs/arvo/

[1]: https://urbit.org/docs/

[timClicks]

Also a decent portion of that address space is also reserved by an individual who doesn't believe that humans are equal. I think he just wants to be king of his own universe. The technology of Urbit is neat, but I find the politics abhorrent.