Joe Bryan (@joemfb) profile picture
~master-morzod
Joined August 2011
Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)Photo by Joe Bryan (@joemfb)
Joe Bryan retweeted
If we want to stop political violence like what happened to Charlie Kirk, we have to be honest about the people who are celebrating it and the people who are financing it. My closing remarks on today's episode of the Charlie Kirk Show:
12,310
36,040
2,753
196,352
6,436,385
America will never be the same.
15,457
87,313
12,291
718,797
88,912,148
11
75
4
1,184
33,464
Joe Bryan retweeted
"Because Azimuth points are the same width as IP addresses,anyone can build a secure overlay Internet on Azimuth...Nothing is wrong with this trivial Unix/Internet/Azimuth stack. We think it's a completely valid product. We even hope someone builds it. But it's not why we came to the party. It feels not wrong, but lame."
1
7
2
42
6,089
Urbit is making a huge leap in networking. Watch @joemfb at Lake Summit as he shows how our team cuts a 130MB file transfer from 26 minutes to 2.8 seconds. This is Urbit's upcoming directed messaging.
2
7
46
2,290
Interesting effect of Orthogonal Persistence: it completely solves short range persistence issues, so mid to long range issues take the front stage. Low-level languages will persistently corrupt data. Static languages will make Schema Upgrade hell. Partial code edits... ouch.
1
1
1
11
1,093
here's the clip
66
450
210
3,942
442,550
For Father's Day I want to share one of the great father/son stories of all time, David Quammen's "Walking Out" from 1980. It originally comes from Quammen's collection "Blood Line" aptly subtitled Stories of Fathers and Sons.
6
26
7
257
85,982
“science as philosophy” tale: a man builds a thermometer. wires it to detect heat. programs it so that when its hot, it makes a noise. one day the thermometer turns to him and says: its strange that things are only “hot” because of how i am programmed. heat exists because of me.
It's strange that beauty is entirely subjective. These are beautiful to us because of how we're programmed. Objectively, they're no more beautiful than any other arrangement of atoms and photons. The day humans go extinct, everything we find beautiful will cease to be beautiful.
35
232
16
2,149
102,370
This represents a variant of the LockStep packet authentication designed for "maturity transformation" from one packet size to another. I'm pleasantly surprised at how well this scheme seems to work. It preserves packet-level authentication in a scale-invariant way, enabling untrusted packet relays that compose well, with minimal overhead on CPU, RAM, and network bandwidth. A publisher ship's runtime will use this to request a "jumbo frame" large packet from Arvo, then generate 1kB UDP packets from it. A client ship's runtime will use this to receive many 1kB packets and assemble them back into a jumbo frame, likely on the order of 8-16MB. The basic idea is that each packet includes a data fragment and the hashes for a pair of nodes further along in the tree. Each pair of hashes can be authenticated immediately in the normal Merkle way: by checking that if you hash the two hashes concatenated to each other, you get the hash at the tree node just above the two nodes. @lukechampine's original LockStep protocol did this. The new trick is to send these future hashes according to a deterministic schedule that minimizes lookahead. It exhibits a very pretty scale invariance property. Before sending any data fragments, a Merkle proof of the first (leftmost) data fragment is sent. These are the tree nodes circled in red in the diagram. Fortunately for our data sizes this initial proof will always fit in a single UDP packet. Now we have a spine of hashes tracing from the root down and to the left to the first leaf. On a client, once we have the initial Merkle proof, we request and receive the first data fragment. This fragment includes hashes for the third and fourth data fragments, which we can verify against their containing hash, which was in the initial Merkle proof. This is the first short horizontal line in the diagram. The second data fragment (first blue line, just under the first black line) contains hashes for a pair of nodes up one layer in the tree, where each node has two leaves under it. This pair of nodes is just under a node in the initial Merkle proof, so it can also be verified immediately. The third data fragment contains hashes for the fifth and sixth data fragments. This is the second short black horizontal line in the diagram. Note the pattern of the short black lines skipping one leaf then providing the next two data fragment hashes, and there's a black line like this for every other data fragment. Now the future pair is validated against the hashes sent in the blue line, not the initial Merkle proof. We've been extending the Merkle spine downward incrementally. The pattern continues. Every other fragment is a black line containing future leaf hashes. The other ones contain hashes of future nodes that are higher up in the tree. This continues until we run out of future nodes that need to be hashes, in which case there are no future hashes for that packet, indicated by an X in the diagram. The way to know which pair of hashes to send in a higher layer can be determined by looking at which subtree is completed by this data fragment, and including the next pair of hashes at the same layer as that subtree. The second data fragment includes hashes one layer up from the leaves because it completes the subtree of the first two leaves and the one node that contains them, which is at the same layer. By only emitting future hashes at the same layer in this way, we minimize how far ahead the hashes can go, and this scheme becomes scale-invariant. Adding a new layer to the tree introduces a new set of pairs of smaller leaves. For each of these pairs, the first packet contains the hashes for the next pair; the second contains the same pair of hashes that would have been emitted if the tree had one fewer layer and this was the containing node, which was a leaf in the original shallower tree. In other words, the traversal at a higher layer is not affected by adding a lower layer. It's still new, but @joemfb and I have tentatively determined that this has the following properties: - stateless client (small->big packets) - publisher (big->small packets) with lookahead window of 1 big packet (just 1! surprising result) - client-side download resumption in O(log n) packet re-sends, where those packets will likely be in cache on the publisher - deterministic schedule for which future hashes will be sent along with a data fragment
What is this? Wrong answers only
1
4
3
27
3,734
Joe Bryan retweeted
We found a secret german data integrity technology
1
1
17
715
Replying to @lukechampine
@lukechampine and @joemfb are building an interesting system in the new networking stack that can take in packets of one size and emit packets of a different size, preserving authentication. We've started to call it "maturity transformation".
2
2
1
10
1,092
I don't know what the end user programming language of the future will look like, but I know it will be called Excel. (apologies to Tony Hoare)
This quoted post is unavailable.
3
25
2,370
Urbit's goal was always, going back to the seminal post we both cited, to rectify the discipline of software engineering and to allow for the possibility of zero-maintenance programs. Peer-to-peer computing falls out of that as an obvious prerequisite, not a final end.
Replying to @tillertolbus
Different goals for different folks. Urbit has two goals: 1. reinvent the stack 2. build a p2p OS Kinode has one goal: 1. build a p2p OS If it is necessary to reinvent the stack to build a p2p OS, Kinode will fail. Otherwise, Urbit risks failure by dividing its attention.
1
5
2
33
2,350
Any creative effort must be the coherent and uncompromising vision of a single mind. Involving many people is useful (later) for implementation, but catastrophic for definition.
7
6
2
44
5,295
Rewriting the same program over and over is a superpower. I've rewritten the same Lisp interpreter about 10 times in as many years, and finally it's in a shape that I like.
14
14
6
187
14,887
"What is Urbit?" It's a network of personal servers. "There are a bunch of things like that now, what makes Urbit different?" Urbit is something the world has never seen: universal computational foundations. Come with me and see how deep this rabbit hole goes. A thread ->
2
38
8
119
16,275
Today I present you with a very special Christmas treat, maybe the best Christmas short story of all time, the high water mark of 20th Century American Naturalism: Lawrence Sargent Hall's 1959 classic, "The Ledge." Hall’s literary career began auspiciously, first with the publication of “The Ledge,” which was initially rejected by both Esquire and The New Yorker but would eventually win the O. Henry Prize in 1960, followed by his debut novel Stowaway which won The William Faulkner Award in 1962. After this, however, Hall’s career fell into relative obscurity. He moved away from writing fiction and concentrated his efforts on literary criticism. He was a celebrated English professor of 40 years at his alma mater, Bowdoin College, and a mainstay of Maine’s literary culture, but otherwise an afterthought in American letters. “The Ledge” remains the crowning achievement of his life’s work. It is short, straightforward, with subtle nods to Hall’s literary forebears, Nathaniel Hawthorne in particular. It tells the story of an unnamed man, his son, nephew, and their Christmas morning duck hunt on Casco Bay. It is the story of man’s hubris, his folly, and his dumb helplessness in the face of nature’s excruciating indifference. “The Ledge” will bring you to your knees. You will feel the chill of the winter Maine weather in your bones. This is a hard read, and for anyone with a son, I recommend caution. That it takes place on Christmas of all days, a fact not dwelled on much in the story, adds a brutal touch to the narrative, and, in the end, is what gives the story its meaning. Screenshots of the story in the tweets to follow. Enjoy.
5
14
6
109
30,726
Joe Bryan retweeted
Urbit engineers are hard at work making Urbit not only personal and durable, but also secure. Watch security expert @rikardhjort discuss with Zorp CEO Logan Allen (@ZorpZK), Evertas Insurance CSO Ryan Lackey (@octal), and UF Engineer @joemfb. youtube.com/watch?v=Sh4JATbX…
2
18
2,151