@@ -1,62 +1,68 @@
11+++
22title = "Toward a New %clay"
3−date = 2016-07-14
4−author = "Curtis Yarvin"
3+date = "2016-07-13"
54description = "Urbit's revision-control system, %clay, is itself due for a (medium-sized) revision!"
6−[taxonomies]
7−posts = ["Essays"]
5+aliases = [
6+ "/posts/essays/toward-a-new-clay",
7+ "/posts/toward-a-new-clay",
8+ "/blog/clay"
9+]
10+
11+[extra]
12+author = "Curtis Yarvin"
13+ship = "~sorreg-namtyv"
814+++
915
1016TLDR: (a) Urbit's revision-control system, `%clay`, is itself due
11−for a (medium-sized) revision! (b) this time, you yourself can
12−really help! (c) please don't give away the buried spoiler!
17+for a (medium-sized) revision! (b) this time, you yourself can
18+really help! (c) please don't give away the buried spoiler!
1319
1420This explanation does not (or at least should not) require you to
1521know anything about Urbit.
1622
1723### But %clay is already pretty cool
1824
19−Of course, nothing in Urbit is allowed to be _lame_. `%clay` is
20−actually quite cool. It does need some work. But first, let's
25+Of course, nothing in Urbit is allowed to be _lame_. `%clay` is
26+actually quite cool. It does need some work. But first, let's
2127talk about what it is and why it's cool.
2228
2329`%clay` is the Arvo vane (Urbit kernel module) for revision
24−control. It's an authenticated, human-memorable global immutable
25−namespace which is also a typed revision-control system. It's
26−also very good at subscription and synchronization. Urbit
30+control. It's an authenticated, human-memorable global immutable
31+namespace which is also a typed revision-control system. It's
32+also very good at subscription and synchronization. Urbit
2733currently uses `%clay` to distribute live hotpatches; we
2834transparently upgrade the whole network with one checkin.
2935
3036### Compared to other revision-control systems
3137
32−`%clay` is clearly a vanilla DVCS in the modern `git` style. Of
38+`%clay` is clearly a vanilla DVCS in the modern `git` style. Of
3339deployed DVCS these days, `%clay` is probably closest in spirit
3440to `bzr` or maybe `fossil`.
3541
3642Obviously, as a part of a larger whole, you would expect any
37−Urbit RCS to be brutally spartan. `%clay` is. It's even more
43+Urbit RCS to be brutally spartan. `%clay` is. It's even more
3844spartan than `bzr`.
3945
40−(In fact, it's really too spartan to fully use at the moment. We
46+(In fact, it's really too spartan to fully use at the moment. We
4147ourselves still use `%clay` mainly as a distribution system, and
42−do all internal development on Github. This is because the
48+do all internal development on GitHub. This is because the
4349present `%clay` lacks a few little conveniences, like commit
44−messages. See upgrades, below.)
50+messages. See upgrades, below.)
4551
46−There are lots of normal DVCS things in `%clay`. But there are
47−also cool and unusual ones. Let's take a brief tour of the
52+There are lots of normal DVCS things in `%clay`. But there are
53+also cool and unusual ones. Let's take a brief tour of the
4854high-level system, then describe what we think should change
4955about it.
5056
5157### Anatomy of a DVCS
5258
53−A DVCS is a historical state graph. A point in this graph is a
59+A DVCS is a historical state graph. A point in this graph is a
5460snapshot of some tree-structured system, like a filesystem.
5561
5662These snapshots, including recursive internal nodes, are stored in
57−a content-addressable store. Pointers in this store are used to
63+a content-addressable store. Pointers in this store are used to
5864construct a revision graph, whose edges are edit and/or merge
59−actions. Some sort of namespace attaches meaningful names to
65+actions. Some sort of namespace attaches meaningful names to
6066current revisions in this graph.
6167
6268There are six big questions in the design of a DVCS:
@@ -72,16 +78,16 @@ There are six big questions in the design of a DVCS:
7278
7379A good way to summarize `%clay` (at present) is that its answers
7480to the first four questions are interesting, and its answers to
75−the last two are boring. We'll talk about all six, though.
81+the last two are boring. We'll talk about all six, though.
7682
7783### Global namespace: [ship desk case]
7884
79−The permanent global name of a `%clay` snapshot is a `++beak`. A beak
85+The permanent global name of a `%clay` snapshot is a `++beak`. A beak
8086is a triple: `[ship desk case]`.
8187
82−The `++ship` is an Urbit ship. A ship is a network address, a
83−human-memorable name and a cryptographic identity. The address
84−width is variable; shorter addresses are more desirable. Most
88+The `++ship` is an Urbit ship. A ship is a network address, a
89+human-memorable name and a cryptographic identity. The address
90+width is variable; shorter addresses are more desirable. Most
8591users have 32-bit addresses, which we map to four-syllable
8692synthetic names like `~sorreg-namtyv`.
8793
@@ -94,44 +100,44 @@ In the Urbit path syntax, the beak is the first three segments of
94100the path: `/ship/desk/case/my/cat/picture`.
95101
96102In most cases, a path parser has a default ship, desk and case,
97−each of which can be replaced by `=`. So `/=home=/my/cat/picture`
103+each of which can be replaced by `=`. So `/=home=/my/cat/picture`
98104is the data at `/my/cat/picture` on your own ship, in the `%home`
99−desk, right now. Immutability doesn't have to be inconvenient.
105+desk, right now. Immutability doesn't have to be inconvenient.
100106
101−`%clay` has no concept of a mutable name or a magic `HEAD`. Use
102−the current date. If you're referring to remote data and you're
107+`%clay` has no concept of a mutable name or a magic `HEAD`. Use
108+the current date. If you're referring to remote data and you're
103109worried about clock skew, subtract a second or three, or be
104−prepared to block. (Or keep your clock unskewed, which shouldn't
105−be _that_ hard.) It's not like mutable-name schemes work any
110+prepared to block. (Or keep your clock unskewed, which shouldn't
111+be _that_ hard.) It's not like mutable-name schemes work any
106112better than this anyway.
107113
108−(Immutability is central to `%clay`. `git` merely adopted the
109−reflog. `%clay` was born in it, molded by it...)
114+(Immutability is central to `%clay`. `git` merely adopted the
115+reflog. `%clay` was born in it, molded by it...)
110116
111117`%clay` is objective; each Urbit node hosts only one ship, but
112−treats domestic and foreign data (almost) the same way. If you
118+treats domestic and foreign data (almost) the same way. If you
113119ask your `%clay` repository for someone else's data, it will
114120block the request until it gets the data, which it then caches.
115121At present, we always fulfill requests from the origin ship,
116122though in future we could certainly use it as a tracker.
117123
118124If `%clay` discards cached data, it keeps a hash, so that your
119−urbit can't violate the local illusion of immutability. We
125+urbit can't violate the local illusion of immutability. We
120126can't, of course, prevent anyone else's ship from sending
121−conflicting claims in different directions. But we can prevent
127+conflicting claims in different directions. But we can prevent
122128programs on our own ship from hearing about them, meaning that we
123129retain local referential transparency.
124130
125131### Local structure: dome, arch, leaf and mark
126132
127−A `%clay` snapshot is called a `++dome`. A dome is the `%clay`
133+A `%clay` snapshot is called a `++dome`. A dome is the `%clay`
128134equivalent of a tree-structured filesystem.
129135
130136Each node in this tree is one uniform structure, an `++arch`,
131−which is both file and directory. A `++leaf` or data value in
137+which is both file and directory. A `++leaf` or data value in
132138this arch is a typed document.
133139
134−A logical document type is called a `++mark`. A mark is a
140+A logical document type is called a `++mark`. A mark is a
135141symbolic pointer to a Hoon source file, in the same dome, for an
136142engine that can validate, compare, and/or convert
137143documents of that mark.
@@ -140,8 +146,8 @@ Let's go over arches and marks a little more closely...
140146
141147#### An arch is a uniform filesystem node
142148
143−Most filesystems and RCS have modal nodes. For instance, a git
144−data node can be a blob or a tree. `%clay` (and the Urbit
149+Most filesystems and RCS have modal nodes. For instance, a git
150+data node can be a blob or a tree. `%clay` (and the Urbit
145151namespace in general) has one kind of node, an `arch`:
146152
147153```
@@ -150,24 +156,24 @@ namespace in general) has one kind of node, an `arch`:
150156```
151157
152158An arch has an optional typed-data leaf, and optional named
153−children. In Unix terms, it's a file, a directory, or both. It
159+children. In Unix terms, it's a file, a directory, or both. It
154160can't be neither: an empty arch, with no data and no children, is
155−auto-pruned. As in `git` (but not in Unix), the hierarchy is a
156−pure function of its leaves. Paths exist only because of the
161+auto-pruned. As in `git` (but not in Unix), the hierarchy is a
162+pure function of its leaves. Paths exist only because of the
157163data they point to.
158164
159165The `++knot` names in an arch path are also a little special.
160166They're typed, in a sense; any knot can be parsed invertibly into
161−a typed atom. The atom is an arbitrary unsigned integer. Its
167+a typed atom. The atom is an arbitrary unsigned integer. Its
162168type (an `++aura` if you know Hoon) defines the semantics and
163169presentation of the atom: for instance, a symbol, a decimal,
164170a date, etc.
165171
166172Ultimately a filesystem is a database, and a path is a primary
167−key. It's very useful to be able to store simple data in this
173+key. It's very useful to be able to store simple data in this
168174key, in a standard and unambiguous way.
169175
170−Why uniform nodes? One, they map nicely to information trees
176+Why uniform nodes? One, they map nicely to information trees
171177where a subtree may need some custom collective presentation,
172178not an automatically generated Apache style directory listing.
173179Two, non-uniform nodes introduce a whole pointless category of
@@ -175,34 +181,34 @@ node type conflicts -- for which life is far too short.
175181
176182#### A mark is an executable document type
177183
178−A mark is "like an executable MIME type." It's best defined as a
184+A mark is "like an executable MIME type." It's best defined as a
179185symbolic mapping from a filesystem to a schema engine.
180186
181187If you have a MIME type `text/html`, you need to watch the
182−Internets to track what the W3C thinks HTML is. If you have an
188+Internets to track what the W3C thinks HTML is. If you have an
183189Urbit mark `%html`, you can functionally construct (using our
184190insanely powerful functional build vane `%ford`) a core which
185191can validate or normalize HTML, apply patch and diff logic, and
186−translate to and from other marks. `%ford` can even reason about
192+translate to and from other marks. `%ford` can even reason about
187193translation paths and construct multistep conversions.
188194
189195(Urbit does not actually have this super-smart `%html` mark right
190−now. But it probably should, huh. And this page was composed in
196+now. But it probably should, huh. And this page was composed in
191197Markdown and translated to HTML by `%ford`.)
192198
193−Where does the source code of a mark live? Within the same dome
194−as the data it describes. `%html` files are described by the
199+Where does the source code of a mark live? Within the same dome
200+as the data it describes. `%html` files are described by the
195201Hoon source at `/===/mar/html`, which perhaps gets mapped into
196202your Unix filesystem at `$ship/home/mar/html.hoon`.
197203
198−This Hoon file is not self-contained. It has dependencies. The
204+This Hoon file is not self-contained. It has dependencies. The
199205`%html` mark may depend on various structures and libraries
200−elsewhere in Urbit. But the source code for all them is within
206+elsewhere in Urbit. But the source code for all them is within
201207this dome.
202208
203209One constraint this puts on domes is that _all domes are
204−descendants of the base system dome_. Domes are versions of one
205−reality, not fragments of that reality. You can't create a new
210+descendants of the base system dome_. Domes are versions of one
211+reality, not fragments of that reality. You can't create a new
206212empty dome -- you'd have no mark source, and be unable to install
207213any validated files.
208214
@@ -213,13 +219,13 @@ weird collisions between non-parallel things don't happen.
213219
214220### Unix mapping and mounting
215221
216−Every DVCS is its own filesystem in a sense. But every DVCS
222+Every DVCS is its own filesystem in a sense. But every DVCS
217223works by mapping and mounting itself, in some sense, to Unix.
218224
219225#### Mapping arches to Unix
220226
221−How do we map arches into Unix? The arch structure maps 1:1 to
222−the Unix directory tree. When the arch has a leaf, we derive a
227+How do we map arches into Unix? The arch structure maps 1:1 to
228+the Unix directory tree. When the arch has a leaf, we derive a
223229dot extension from the file's mark.
224230
225231So, supposing we have a leaf `foo` which is an HTML document
@@ -233,19 +239,19 @@ foo/moo.html
233239```
234240
235241This maps invertibly back into `%clay`, and it looks perfectly
236−unremarkable from the Unix side. File completion sometimes gets
242+unremarkable from the Unix side. File completion sometimes gets
237243a little annoying.
238244
239245It would be if nice every possible Unix tree, even every
240246plausible Unix tree, mapped cleanly and completely to a `%clay`
241−dome. There are always horrid exceptions, even in common use --
242−like extensionless Makefiles. The trivial solution is that if
247+dome. There are always horrid exceptions, even in common use --
248+like extensionless Makefiles. The trivial solution is that if
243249`%clay` finds a file mysterious, it won't track it.
244250
245251One often unquestioned assumption in new system software is that
246−compatibility with old content is important. It's nice, but not
247−that important. Most content in a new system is new content.
248−The important compatibility direction is export, not import. So
252+compatibility with old content is important. It's nice, but not
253+that important. Most content in a new system is new content.
254+The important compatibility direction is export, not import. So
249255the new system should subset, not superset, the ancient formats.
250256
251257### Mounting desks to Unix
@@ -256,36 +262,36 @@ the Unix directory of the same name within the pier.
256262When files in a mount directory change, `%clay` notices it
257263through `libuv` filesystem watching (ultimately `inotify()` or
258264`FSEvents`), computes diffs, and automatically commits them to
259−the desk. And if `%clay` creates a new version for its own
265+the desk. And if `%clay` creates a new version for its own
260266urbity reasons, it will revise the Unix files.
261267
262268Keeping all mounted files in the same Unix subdirectory makes it
263−easy to move the whole pier in a tarfile. Edits applied while
269+easy to move the whole pier in a tarfile. Edits applied while
264270the `urbit` process isn't running will be detected and applied
265271the next time it starts.
266272
267273In conventional DVCS terms, `%clay` has a 1:1 static mapping from
268−branches to working directories. Why?
274+branches to working directories. Why?
269275
270−Why not? Do we need that "what branches are checked out in what
271−directories" piece of state? Why force the user to worry about
272−which branches are mounted to which directories? Why not just
273−mount them under their own names? Is there some crippling global
276+Why not? Do we need that "what branches are checked out in what
277+directories" piece of state? Why force the user to worry about
278+which branches are mounted to which directories? Why not just
279+mount them under their own names? Is there some crippling global
274280shortage of disk space for source files? Our planet is running
275−out of inodes? Wut?
281+out of inodes? Wut?
276282
277283### Queries, subscriptions and synchronization
278284
279285There are two ways to get data out of `%clay`: dereference the Urbit
280−namespace, or send a request/subscription. Sync is in userspace.
286+namespace, or send a request/subscription. Sync is in userspace.
281287
282288#### Dereferencing the namespace
283289
284−`%clay` is just part of the full Urbit namespace. The full namespace
290+`%clay` is just part of the full Urbit namespace. The full namespace
285291has a two-character prefix which expresses what vane to query
286−(always `%c` for `%clay`). The suffix is a mode flag; `%x` for
292+(always `%c` for `%clay`). The suffix is a mode flag; `%x` for
287293data, `%y` for a list of children, `%z` for the whole arch
288−subtree, `%w` for the version number. (Each vane has its own
294+subtree, `%w` for the version number. (Each vane has its own
289295modes.)
290296
291297For example, the full path to my cat picture is
@@ -299,6 +305,7 @@ For example, the full path to my cat picture is
299305If the data model is `++image` (hypothetical -- we don't really
300306do image processing in Urbit right now), the normal way to load
301307this data in a Hoon program is
308+
302309```
303310.^(image /cx/~sorreg-namtyv/home/[now]/my/cat/picture)
304311```
@@ -315,15 +322,15 @@ learn it in future, `[~ ~]` if we know we'll never know it, or
315322`[~ ~ value]` if we know it now.
316323
317324In high-level shell programming (`:dojo`), we can make apparently
318−"blocking" requests to `%clay`. What actually happens: when `%clay`
325+"blocking" requests to `%clay`. What actually happens: when `%clay`
319326produces `~`, we stop trying to compute this expression and wait
320−on one or more blocking requests. When we are unblocked, we
327+on one or more blocking requests. When we are unblocked, we
321328recompute again from scratch; the intermediate computation is not
322329saved at all.
323330
324−Urbit doesn't like to block. It likes to save coherent data
325−structures, not incomplete computation contexts. When Urbit
326−pretends to block for you, this is a courtesy service. You
331+Urbit doesn't like to block. It likes to save coherent data
332+structures, not incomplete computation contexts. When Urbit
333+pretends to block for you, this is a courtesy service. You
327334can return the courtesy of not expecting miracles from it.
328335When in doubt, send an actual request.
329336
@@ -334,7 +341,7 @@ namespace, but can (a) read across case (revision) ranges,
334341and (b) filter by subtree.
335342
336343If the revision range in a request stretches into the future, the
337−request becomes a subscription. Urbit messaging always regards a
344+request becomes a subscription. Urbit messaging always regards a
338345single response as a special case of a subscription.
339346
340347Generally, the response stream created by a request starts with
@@ -344,10 +351,10 @@ closed when the request range is complete.
344351
345352#### Userspace synchronization
346353
347−Synchronization per se is done in userspace. In `%clay`, sync just
354+Synchronization per se is done in userspace. In `%clay`, sync just
348355means an application (typically the `%kiln` module in the system
349356app `:hood`) subscribes to a remote desk, and saves its change
350−stream locally. So sync is really an emergent feature of `%clay`.
357+stream locally. So sync is really an emergent feature of `%clay`.
351358
352359### Revision graph
353360
@@ -355,16 +362,16 @@ stream locally. So sync is really an emergent feature of `%clay`.
355362Some changes are proposed below, however.
356363
357364As in any DVCS, a dome is the child of one parent dome (an edit)
358−or more parents (a merge). Often the parent of any revision is
365+or more parents (a merge). Often the parent of any revision is
359366the previous version in the sequence, but it can be anything.
360367
361368### Storage model
362369
363−`%clay`'s storage model is completely uninteresting. It's also a
370+`%clay`'s storage model is completely uninteresting. It's also a
364371little funky and I want to change it, so I'll be super vague.
365372
366373A `++lobe`, a 128-bit atom, is the Merkle hash of a `blob`, which
367−is essentially a leaf. All blobs are stored in a hash table
374+is essentially a leaf. All blobs are stored in a hash table
368375(`(map lobe blob)`), and reference each other through lobes.
369376
370377At present we don't GC this space at all, because we don't
@@ -373,12 +380,12 @@ actually delete any roots.
373380A `blob` may be `%direct`, meaning it's defined as the data
374381itself, or `%delta`, defined as a change against another blob.
375382This is motivated by the desire not to store a full snapshot
376−on every file change. But it may be misguided (see changes #0
383+on every file change. But it may be misguided (see changes #0
377384and #1 below).
378385
379386We cache the full arch tree (`++ankh`) of the latest state in
380−each desk. For current queries, this mitigates a lot of the
381−opportunities for slowness that this data model creates. But the
387+each desk. For current queries, this mitigates a lot of the
388+opportunities for slowness that this data model creates. But the
382389disparity between the speed of the current version and the
383390speed previous version is an undesirable glitch in the API.
384391
@@ -390,38 +397,38 @@ We actually have _three_ motivations for change...
390397
391398At present, we're still using `git` as a development platform.
392399`%clay` is certainly used, but just for its strength --
393−deployment. This is starving Urbit of essential experience and
394−basically keeping it lame. So, change is necessary.
400+deployment. This is starving Urbit of essential experience and
401+basically keeping it lame. So, change is necessary.
395402
396403### Change must come from the community
397404
398−"Why?" you ask? "Why must change come from the community?" Well,
399−one, duh. Change always comes from the community.
405+"Why?" you ask? "Why must change come from the community?" Well,
406+one, duh. Change always comes from the community.
400407
401−But two, a great force for change has departed. Philip Monk,
408+But two, a great force for change has departed. Philip Monk,
402409late maintainer of `%clay` (`~wictuc-folrex` in the old era,
403−`~wicdev-wisryt` in the new), is tired of types and typing. He
410+`~wicdev-wisryt` in the new), is tired of types and typing. He
404411spent September roaming with the bison in Montana, and is now
405−literally on his way to Patagonia. On the one hand, Phil still
412+literally on his way to Patagonia. On the one hand, Phil still
406413exists and is not at all unwilling to answer questions, even dumb
407−questions. On the other hand, _Patagonia_.
414+questions. On the other hand, _Patagonia_.
408415
409−So %clay needs a new owner, or owners. At this point, we think
416+So %clay needs a new owner, or owners. At this point, we think
410417the community might be indoctrinated enough to take a stab.
411418
412419Alas, `%clay` already has a pretty checkered ownership history.
413420Long ago I built a `%clay` that was really the only serious
414−design failure in Arvo. Then Jared Hance, a brilliant summer
421+design failure in Arvo. Then Jared Hance, a brilliant summer
415422intern, took my broken `%clay` and reformed it into a minimal `git`
416−in about a week. Then Philip modified Jared's `%clay` to be a
423+in about a week. Then Philip modified Jared's `%clay` to be a
417424_typed_ revision-control system, which is a pretty cool thing.
418425
419426But... in none of these transitions did `%clay` actually get
420−_rewritten_ per se. While it's a reasonably good vane, it
427+_rewritten_ per se. While it's a reasonably good vane, it
421428does contain one or two historical decisions.
422429
423430We know a lot more about Arvo than we used to, and I think we can
424−revisit many of these decisions. The point of this post is to
431+revisit many of these decisions. The point of this post is to
425432start a `%clay` design discussion that will evolve into actual
426433collective action.
427434
@@ -431,41 +438,41 @@ We're planning a rewards program for contributions to `%clay`.
431438It'll simply involve setting a time period and a pool of real
432439estate, and allocating the pool by judged contribution.
433440
434−This is an experimental program. We'll say more about it later.
435−It may not even happen. It may happen very differently. When in
441+This is an experimental program. We'll say more about it later.
442+It may not even happen. It may happen very differently. When in
436443doubt, don't talk about Fight Club.
437444
438445## Some openings for change
439446
440−Change, of course, must come from the community. No one can
441−dictate change from the top down. We _know_ this. But...
447+Change, of course, must come from the community. No one can
448+dictate change from the top down. We _know_ this. But...
442449
443450This list of suggested changes for the next `%clay` is not in any
444−sort of order: priority, structure, difficulty, etc. Let's talk
445−about the changes first; make actual plans second. The numbers
451+sort of order: priority, structure, difficulty, etc. Let's talk
452+about the changes first; make actual plans second. The numbers
446453here are just for convenience.
447454
448−*All changes in this program will happen on the next-generation
449−(`cc-release`) branch.*
455+_All changes in this program will happen on the next-generation
456+(`cc-release`) branch._
450457
451458### Change #0: state reform
452459
453460As mentioned above, the storage model in `%clay`, as the code is
454−at present, feels a little funky to me. There's nothing terribly
461+at present, feels a little funky to me. There's nothing terribly
455462awful about it, but it doesn't feel quite right and its scaling
456463properties are very questionable.
457464
458465Briefly, `%clay` today actually derives the hierarchy from its
459−leaves -- rather than _logically_ so deriving it. Only leaves
460−are in the object store. This is just weird, and you can feel
461−`%clay` trying to be something other than what it is. (I blame
466+leaves -- rather than _logically_ so deriving it. Only leaves
467+are in the object store. This is just weird, and you can feel
468+`%clay` trying to be something other than what it is. (I blame
462469Jared, though it's actually my fault.)
463470
464471To store a dome snapshot (`++yaki`), the present code keeps a
465−`(map path blob)` at the root of the dome. This is enough to
472+`(map path blob)` at the root of the dome. This is enough to
466473reconstruct the `++arch` structures the namespace exports.
467474However, all kinds of basic arch operations, like basically `ls`,
468−become `O(n)` in the number of leaves in the dome. This is
475+become `O(n)` in the number of leaves in the dome. This is
469476obviously not quite right.
470477
471478I believe the storage structure should be (removing everything
@@ -486,46 +493,46 @@ Here a `++root` is the universal node, with children addressed by
486493hash, but logically representing the abstract `++arch`.
487494
488495Users of hierarchical filesystems basically expect performance
489−qualities that match the filesystem structure. A huge directory
490−in one place shouldn't affect search speeds somewhere else. And
496+qualities that match the filesystem structure. A huge directory
497+in one place shouldn't affect search speeds somewhere else. And
491498of course, `ls` can't be O(n) in the size of the filesystem.
492499
493500### Change #1: store edges, cache snapshots
494501
495−The current `%clay` is snapshot-oriented, like `git`. The actual
496−changes are discarded. I think this is a bad choice.
502+The current `%clay` is snapshot-oriented, like `git`. The actual
503+changes are discarded. I think this is a bad choice.
497504
498505`%clay` is not a research project, and has no idea of going as
499−far as `darcs` in the "set-of-patches" direction. However, at
506+far as `darcs` in the "set-of-patches" direction. However, at
500507present we actually throw away the patch (`++nori`).
501508
502509I think the right DVCS design for `%clay` is a snapshot-oriented
503−system which also keeps the patches. The revision graph has the
504−same general shape as in the old `git` style. But rather than
510+system which also keeps the patches. The revision graph has the
511+same general shape as in the old `git` style. But rather than
505512just deriving the diffs from the snapshots, we can also derive
506−the snapshots from the diffs. When in doubt, we save space by
513+the snapshots from the diffs. When in doubt, we save space by
507514dropping the snapshot, not the diff.
508515
509516For an RCS which is also a general-purpose typed filesystem, it's
510517clear that considering edges (deltas, diffs, changes, etc) as the
511−primary state is optimal. You can use diff functions to derive
518+primary state is optimal. You can use diff functions to derive
512519deltas from snapshots, but in many data models this will always
513−risk losing information from the original action. When in
520+risk losing information from the original action. When in
514521doubt, save what the user did.
515522
516523This is even more true for operations on the tree structure, not
517−on individual leaves. Consider the heuristic energy that `git` has
524+on individual leaves. Consider the heuristic energy that `git` has
518525to go through to figure out when `mv` has been used -- even when
519−`git mv` has been used! (According to the `git` faq, this is
526+`git mv` has been used! (According to the `git` faq, this is
520527because the lowest common denominator remains Linus' style of
521528working with emailed patches.)
522529
523−Edges probably don't even need to be deduplicated. They also can
530+Edges probably don't even need to be deduplicated. They also can
524531contain metadata such as commit messages (which `%clay` has no
525532place for at present).
526533
527534Once edges are recognized as unreclaimed primary state, snapshots
528−can always be derived from the edges. So we can reclaim memory
535+can always be derived from the edges. So we can reclaim memory
529536being used for snapshots.
530537
531538It's also straightforward to have a kind of edge which is a list
@@ -533,27 +540,27 @@ of edges, letting us collapse feature branches trivially.
533540
534541Finally, I think references in edges should wherever possible
535542indirect through the namespace, rather than directly to the
536−content address. The user has strictly more information if you
543+content address. The user has strictly more information if you
537544say you merged `/~sorreg-namtyv/base/987/`, than if all you have
538545is the hash that this reference refers to.
539546
540547### Change #2: unconfuse merge and rebase
541548
542549The dispute between merge and rebase has all the bad smells of
543−a historical wart. But here I'm just not sure what's right.
550+a historical wart. But here I'm just not sure what's right.
544551
545552The user has one _intent_ whether merging or rebasing, which is
546−to combine changes on one desk into another. The _algorithm_
553+to combine changes on one desk into another. The _algorithm_
547554used to construct this change, whether operational transformation
548555(rebasing) or a 3-way diff (merging) has one goal: use all
549556available information to model the authors' editing intent as
550−realistically as possible. If we save edges, we should use
557+realistically as possible. If we save edges, we should use
551558edges.
552559
553560Whether to define the resulting change graph with the typical
554561structure of a merge, or the typical structure of a rebase,
555562strikes me as quite arbitrary, and well into the land of "pick
556−one and force everyone to stick with it." But this call demands
563+one and force everyone to stick with it." But this call demands
557564a real expert on RCS workflow, which I'm not.
558565
559566Since `%clay` is a humble part of a larger system and needs to
@@ -563,23 +570,23 @@ different preferences" is not usually the right answer.
563570### Change #3: comment jihad
564571
565572`%clay` needs to be brought up to the documentation standards of
566−`%jael`. This means (a) the formatting matches; (b) the comments
573+`%jael`. This means (a) the formatting matches; (b) the comments
567574are at the same level of quality and ubiquity.
568575
569576The basic commenting standards are: (a) every core (class) needs
570577a block comment explanation; (b) every gate (function) needs a
571578line comment; (c) every face (variable) needs a line comment.
572579
573−This is obviously not just a clerical task. To write these
574−comments, you need to actually understand the code. This is a
580+This is obviously not just a clerical task. To write these
581+comments, you need to actually understand the code. This is a
575582nontrivial but extremely rewarding task, even if the code you're
576583commenting will get replaced.
577584
578585Also, the _symbols_ in `%clay` must be brought up to `%jael`
579−quality. There is still far too much four-letterism here.
586+quality. There is still far too much four-letterism here.
580587
581588NB: `%clay` is about 3000 lines of pretty clean Hoon, which is
582−certainly not at all undocumented. We're not exactly looking at
589+certainly not at all undocumented. We're not exactly looking at
583590a Matterhorn of reverse engineering here.
584591
585592### Change #4: knot to coin internal paths
@@ -590,26 +597,26 @@ ASCII symbol as an atom) must be within the range of the
590597invertible noun printer, `++co`.
591598
592599`++co` maps a described data value, a `++coin`, to an ASCII
593−string which is a valid URL segment and Unix filename. Its
600+string which is a valid URL segment and Unix filename. Its
594601friend `++so` inverts the function, perfectly.
595602
596−Almost every knot is the simplest kind of `coin`, a `dime.` A
603+Almost every knot is the simplest kind of `coin`, a `dime.` A
597604dime is a single atom with a rendering type (`aura` if you know
598−Hoon). Auras include `@tas` for symbolic text, `@ud` for
599−unsigned decimal, `@p` for Urbit ship, etc. `++co` can also
605+Hoon). Auras include `@tas` for symbolic text, `@ud` for
606+unsigned decimal, `@p` for Urbit ship, etc. `++co` can also
600607print arbitrary cells, though less prettily.
601608
602609This feature is cool because it lets us stuff data into names in
603−a precise and predictable way. Usually the names in a namespace
610+a precise and predictable way. Usually the names in a namespace
604611are human-meaningful and human-originated, and are therefore
605−symbolic, but there are exceptions. For example, blogposts may
612+symbolic, but there are exceptions. For example, blogposts may
606613want to be named by date.
607614
608615However, it is slightly retarded for the fundamental state of our
609616arches to store the _rendered_ string (`++knot`), not the source
610−data (`++coin`). It leads to all kinds of inefficiencies where
611−we're doing way more parsing than we want to. (Parsing also
612−should be jet-propelled, but is not.) It also makes it way too
617+data (`++coin`). It leads to all kinds of inefficiencies where
618+we're doing way more parsing than we want to. (Parsing also
619+should be jet-propelled, but is not.) It also makes it way too
613620easy to sneak in bad data and make it `%clay`'s problem.
614621
615622So, logically, we'd be shifting from
@@ -624,7 +631,7 @@ So, logically, we'd be shifting from
624631
625632Also, a vanilla `++map` might not be quite the right data
626633structure at this point -- we probably want a custom sorted tree,
627−which separates auras and imposes a custom order on each. This
634+which separates auras and imposes a custom order on each. This
628635would let us serve simple range queries quite efficiently, so
629636long as they match the standard aura order (eg, alphabetical
630637order for symbols).
@@ -632,23 +639,23 @@ order for symbols).
632639### Change #5: dome semantics by reference
633640
634641While the principle that a dome contains its own mark source is
635−elegant, it's a pain in the ass to implement properly. `%clay`
642+elegant, it's a pain in the ass to implement properly. `%clay`
636643certainly does not do it justice at present.
637644
638645The complexity is in checking every change to see whether it
639646includes a change to the mark source (which is very unlikely),
640647or a dependency of the mark source (slightly less unlikely).
641648And if so, do we have to update all the leaves in the dome?
642−This is also very unlikely. And yet, it happens.
649+This is also very unlikely. And yet, it happens.
643650
644651Instead, we could simply designate, by beak (ship, desk, and
645−case) a dome which contains the canonical mark source. This
652+case) a dome which contains the canonical mark source. This
646653would rarely change -- you could upgrade a desk to a new mark
647654dome, but it would be a manual operation and probably a rare one.
648655
649656The common case would just be that everyone uses the current
650657version of the `%base` desk when a desk is created, and never
651−changes it. In practice, this means not every two-byte edit
658+changes it. In practice, this means not every two-byte edit
652659needs to get involved in detecting and/or applying schema
653660evolution, a welcome simplification.
654661
@@ -660,14 +667,14 @@ design is one mark per message structure, the mark namespace gets
660667overloaded fast.
661668
662669So we added an app hierarchy within the mark namespace, using hep
663−`-` as a divider. The `:talk` app expects messages with the mark
670+`-` as a divider. The `:talk` app expects messages with the mark
664671`talk-message`, whose sources is in `/===/mar/talk/message`.
665672
666673This helps, but it doesn't solve another two problems: trivial
667674nouns and containers.
668675
669676It's retarded to have to add a source file to send a `@ud` or a
670−`*`. An aura (see above under `coin`) should be a trivial mark.
677+`*`. An aura (see above under `coin`) should be a trivial mark.
671678
672679Also, we need to be able to construct standard containers of
673680marks: lists, maps, sets, and tuples (or, as a database person
@@ -679,32 +686,32 @@ Another kind of complex mark that needs to exist: conflict,
679686difference, and invalid marks.
680687
681688Conflict data and difference data are two types of noun that are
682−generated in normal RCS operation. If we have the concept of an
689+generated in normal RCS operation. If we have the concept of an
683690`%html` noun, we also need the concept of an invertible change
684691from one `%html` file to another; a merge conflict between two
685692`%html` files; and even a conflict between two marks.
686693
687694There are two kinds of changes to a dome: manual and automatic.
688695Manual changes are initiated by the user and should crash, unless
689−manually overridden. Automatic changes should not crash, ever,
690−and should not lose data. If they create problems, which they
696+manually overridden. Automatic changes should not crash, ever,
697+and should not lose data. If they create problems, which they
691698shouldn't, these problems will need to be cleaned up manually.
692699
693700This is a complex space to navigate, but it requires a conflicted
694701merge to be able to resolve to a valid dome containing conflict
695−leaves, not crash. The present `%clay` goes halfway there and
702+leaves, not crash. The present `%clay` goes halfway there and
696703resolves to a scratch desk, but we can go farther: the whole `git`
697−dumpster-fire of "merge state" deserves destruction. One change
704+dumpster-fire of "merge state" deserves destruction. One change
698705is a merge that produces conflicts, the next is an edit that
699706resolves them.
700707
701−Similarly, all sorts of errors can produce invalid leaves. One
708+Similarly, all sorts of errors can produce invalid leaves. One
702709particularly rare case which is nonetheless important is schema
703710evolution: what happens to files in a mark, when the mark source
704−code changes. They all need to be re-normalized. And this
705−operation can fail, producing an invalid value. (We don't do any
711+code changes. They all need to be re-normalized. And this
712+operation can fail, producing an invalid value. (We don't do any
706713of this now, BTW, although `%ford` has the dependency mechanism to
707−make it possible.) Again, in some cases this should crash the
714+make it possible.) Again, in some cases this should crash the
708715update to the mark source; in other cases, there is no local way
709716to handle the crash.
710717
@@ -715,46 +722,46 @@ into using the bad content.
715722
716723The bottom line: a mark needs to be split into two models:
717724`++mark`, the original symbolic mark with its source engine, and
718−something like `++logo`, the meta-mark or mark container. The
725+something like `++logo`, the meta-mark or mark container. The
719726mark is the building block; the logo is the building.
720727
721728### Change #8: filesystem hierarchy standard
722729
723730While all filesystems remain forks of one ur-filesystem, the
724−standard tends to enforce itself. But not define itself.
731+standard tends to enforce itself. But not define itself.
725732
726−I've always hated the current `/mar` for marks, for instance. I
727−prefer `/mac`. But others hate this. And so on. Also, by
733+I've always hated the current `/mar` for marks, for instance. I
734+prefer `/mac`. But others hate this. And so on. Also, by
728735necessity, our nascent "FHS" is heavily biased toward system
729−data, away from user data. This is retarded.
736+data, away from user data. This is retarded.
730737
731−Please focus all bikeshedding energy on this problem! Better to
738+Please focus all bikeshedding energy on this problem! Better to
732739bikeshed about what color to paint the bikeshed, then what shape
733740the bike gears should be.
734741
735742### Change #9: security and access control
736743
737−Oh, yeah, that! They always say not to design security in at the
738−last minute. Unfortunately, this is exactly what we did. There
739−is no security at all in present `%clay`. (There is also no push
744+Oh, yeah, that! They always say not to design security in at the
745+last minute. Unfortunately, this is exactly what we did. There
746+is no security at all in present `%clay`. (There is also no push
740747operation in `%clay`, so it's not as mad as you think -- anyone
741748can read anything, no one can write anything, which is fine in a
742749young network where there are no secrets.)
743750
744751Fortunately, we don't have anything super fancy planned in this
745−area, just basic access-control lists. These obviously need to
752+area, just basic access-control lists. These obviously need to
746753be designed along with the filesystem-hierarchy standard (see
747754elsewhere).
748755
749−Is it even possible to do ACLs properly in `%clay`? At this
750−point, yes. There is an event-oriented security model in Arvo,
756+Is it even possible to do ACLs properly in `%clay`? At this
757+point, yes. There is an event-oriented security model in Arvo,
751758though it isn't really hooked up to anything at the moment.
752−There is also a privilege badge system in `%jael`. The
759+There is also a privilege badge system in `%jael`. The
753760combination of these two features is quite sufficient for a
754761decent ACL model.
755762
756763The idea of event-based security is that every event passed to
757−`%clay` has a security `++mask`. The mask tells us:
764+`%clay` has a security `++mask`. The mask tells us:
758765
759766- the set of ships that has _contaminated_ this event
760767- the set of ships that has _endorsed_ this event
@@ -762,20 +769,20 @@ The idea of event-based security is that every event passed to
762769
763770To simplify this system a little and maybe explain it, the
764771security question in any event is: (a) who has _caused_ this
765−event? (b) who may receive its _effects_? (These sets never
772+event? (b) who may receive its _effects_? (These sets never
766773include your urbit itself, of course, only neighbors.)
767774
768775For example, if we're doing something because ship X sent us a
769776packet, and we're going to leak the results of this something
770777back in a response to ship X, once `%clay` knows what our ship
771778thinks of its neighbor X, we should have a good picture of what
772−effects this event should be allowed to cause. If `%jael` knows
779+effects this event should be allowed to cause. If `%jael` knows
773780that X is our friend, `%clay` can know that X can see our cat
774781picture.
775782
776783Causality is separated out into contamination (data from ship X
777784got into the event) and endorsement (ship X signed the whole
778−event). (This may be overdesign.) All these sets can be defined
785+event). (This may be overdesign.) All these sets can be defined
779786as the empty set (eg, releasing to the empty set means no output)
780787or the universal set (input from the universal set could be from
781788anyone, ie, is unauthenticated).
@@ -784,11 +791,11 @@ anyone, ie, is unauthenticated).
784791
785792When you read a leaf from a `%clay` dome, `%clay` is happy to
786793perform any necessary mark conversions (the actual work is done
787−by the `%ford` build vane). You can also just ask for the leaf
794+by the `%ford` build vane). You can also just ask for the leaf
788795in whatever mark `%clay` has it.
789796
790797It's confusing to think of this mark extension as part of the
791−path. You shouldn't say `+cat /===/app/talk/hoon`. You should say
798+path. You shouldn't say `+cat /===/app/talk/hoon`. You should say
792799`+cat /===/app/talk`, or maybe `+cat %hoon /===/app/talk`.
793800
794801A mark-extended namespace may be added to `%ford`, to compensate.
@@ -797,7 +804,7 @@ But type-converting reads are the rare case, not the common case.
797804### Change #11: real queries and indexes
798805
799806A filesystem should postpone being a database until it can no
800−longer do so. So, please don't think about this problem yet.
807+longer do so. So, please don't think about this problem yet.
801808I'm joking, of course!
802809
803810### Change #12: meta-query
@@ -806,24 +813,24 @@ Make it possible to query the list of desks in a ship :-)
806813
807814### Change #13: completely redesign Unix mounting
808815
809−This is actually a series of changes, described below. But
816+This is actually a series of changes, described below. But
810817broadly...
811818
812819One of Urbit's experiments is live two-way synchronization
813820between the Unix filesystem and the revision-control store.
814−Dropbox style, as it were. I think this experiment is basically
815−a failure. It may be right for cat pictures, but I don't think
821+Dropbox style, as it were. I think this experiment is basically
822+a failure. It may be right for cat pictures, but I don't think
816823it's right for Urbit.
817824
818825The fundamental problem is that there are all sorts of weird,
819−flaky ways applications use a mutable filesystem. It's not at
826+flaky ways applications use a mutable filesystem. It's not at
820827all unusual for an editor to change a file by deleting the old
821828file, then writing the new one in its place.
822829
823−You really don't want this to produce a delete change. A simple
830+You really don't want this to produce a delete change. A simple
824831cooldown timer for the sync helps a lot with this, but causes
825−other weirdnesses. In theory, a sync can even cause another
826−change to the same version of the same desk. And so on.
832+other weirdnesses. In theory, a sync can even cause another
833+change to the same version of the same desk. And so on.
827834
828835The thing we fear is just too funky to work (there is some
829836strange "reflux bug" still affecting it) is any sort of _two-way_
@@ -834,8 +841,8 @@ cold storage, is that there's no useful workflow that actually
834841uses it -- at least, when editing Urbit code.
835842
836843It's actually really nice to have files on a disk that nothing
837−will automatically modify. If you're using `git` and `%clay` in
838−parallel, it's nice to mount your `git` working directory. The
844+will automatically modify. If you're using `git` and `%clay` in
845+parallel, it's nice to mount your `git` working directory. The
839846original Urbit workflow was to work out of a mounted directory,
840847but the two-way updates were just too weird and scary, so now we
841848tend to edit elsewhere and copy in -- which is lame.
@@ -848,25 +855,25 @@ FUSE (in a somewhat experimental state), and filesystem watching
848855
849856The FUSE interface is a good one, though it needs more work.
850857Moreover, it exports the whole Urbit namespace to Unix, not just
851−`%clay`. For reading Urbit data, FUSE is basically ideal.
858+`%clay`. For reading Urbit data, FUSE is basically ideal.
852859
853−The obvious impulse is to make a read-write FUSE. But I don't
860+The obvious impulse is to make a read-write FUSE. But I don't
854861think FUSE should be used for recording changes to an RCS, even
855862though the FUSE API sends you changes as events.
856863
857−There are three reasons for this. One, the actual filesystem
864+There are three reasons for this. One, the actual filesystem
858865writes that Unix apps make are typically of low quality and
859−confusing. Some crack-brained editor will find it completely
866+confusing. Some crack-brained editor will find it completely
860867normal and unremarkable to modify a byte in a file, by deleting
861−the file and rewriting it from scratch. You can use a cooldown
868+the file and rewriting it from scratch. You can use a cooldown
862869timer to wait for this nonsense to converge, but leaving data in
863870heuristic limbo seems like a very bad idea in FUSE-land.
864871
865872The second is that people derive a lot of comfort from seeing
866−actual files on an actual disk. The idea that your datas are
873+actual files on an actual disk. The idea that your datas are
867874stuck in Urbit is slightly terrifying as a user experience.
868−We claim to be reliable. It's actually hard to imagine datas that
869−couldn't be recovered from either snapshot or log. But how can
875+We claim to be reliable. It's actually hard to imagine datas that
876+couldn't be recovered from either snapshot or log. But how can
870877you possibly believe us on this?
871878
872879So I think it's correct to lean all the weight of reading data
@@ -881,35 +888,35 @@ But this is a separate discussion.
881888
882889### Change #15: one-way, nonced working desk
883890
884−A one-way mount is a much less crazy thing. It works the same
891+A one-way mount is a much less crazy thing. It works the same
885892way our two-way mount works today, but with only one writer: the
886893user, via Unix.
887894
888895Since a mount is one-way, it needs to have an absolute write lock
889−on the desk it's writing to. We cannot have a `%home` desk which
896+on the desk it's writing to. We cannot have a `%home` desk which
890897we are also editing with filesystem watching.
891898
892899Instead, we have a temporary and anonymous state that's connected
893−itself with your `%home` desk. Desk names have to be valid
894−symbols, but we can use `hep`, `-`. A reasonable choice is to
900+itself with your `%home` desk. Desk names have to be valid
901+symbols, but we can use `hep`, `-`. A reasonable choice is to
895902pick a random 32-bit name for each mount instance -- so the desk
896−name is not `%home`, but `%home-bosdut-sigder`. (Randomizing
903+name is not `%home`, but `%home-bosdut-sigder`. (Randomizing
897904instead of always using a fixed suffix, say, `%home-work`,
898905just makes all kinds of screwups less likely.)
899906
900907You create a mount simply by mkdiring the desk name in your pier.
901908Just the base name, Urbit will come up with the `%bosdut-sigder`
902−part and put it in a `.desk` file at the desk root. To forget
903−the mount, just delete this file. This avoids the ugly confusion
909+part and put it in a `.desk` file at the desk root. To forget
910+the mount, just delete this file. This avoids the ugly confusion
904911between unmounting a mount, and deleting all the data in it.
905912
906913(Urbit will keep a set of the mount desks it's used and
907−forgotten, not reuse them, and block all requests into them. By
914+forgotten, not reuse them, and block all requests into them. By
908915blocking, we can delete without losing referential transparency.
909916Yes, Virginia, Urbit can forget!)
910917
911918Note that the auto-reload behavior of applications is extremely
912−useful when testing. When testing, you can either switch the
919+useful when testing. When testing, you can either switch the
913920active application desk from `%home` to `%home-bosdut-sigder`,
914921or activate autosync from the latter to the former.
915922
@@ -932,7 +939,7 @@ Otherwise, you're basically, like, reinventing Perforce.
932939
933940More concretely, the Urbit experience today requires `:hood`
934941pokes (Urbit shell commands, basically) for all nontrivial RCS
935−operations, from mounting to merging. Again, there are lots of
942+operations, from mounting to merging. Again, there are lots of
936943situations in which you can write into a mounted filesystem (the
937944protocol for Unix communication could be easily adapted to
938945communicate with a browser-based editor, for instance), but you
@@ -942,20 +949,20 @@ Even commands like mounting and merging are easily communicated,
942949through filesystem writes that any Unix tool (including many
943950macro editors).
944951
945−This is very easy for a filesystem watcher. One such protocol:
952+This is very easy for a filesystem watcher. One such protocol:
946953if you want to send the daemon a command relative to some
947−directory, write a `.ur-command` metafile in that directory. The
954+directory, write a `.ur-command` metafile in that directory. The
948955daemon accepts the command by deleting it.
949956
950957The user experience is that you configure your mounts by simply
951−making the directories and configuring dotfiles in them. Unlike
958+making the directories and configuring dotfiles in them. Unlike
952959in the present system, Urbit will never modify your datas
953960silently in the background.
954961
955962Here's what this looks like:
956963
957−Suppose you want to edit some data in the `%home` desk. Your
958−data is in `/=home=/mar/talk`. You also want a read-only view
964+Suppose you want to edit some data in the `%home` desk. Your
965+data is in `/=home=/mar/talk`. You also want a read-only view
959966of `/=home=/arvo`.
960967
961968From Unix, you do something like this:
@@ -968,18 +975,18 @@ From Unix, you do something like this:
968975 $ touch home/arvo/.read
969976```
970977
971−Want to commit your changes? That means merging the working
972−desk, `%home-bosdut-sigder`, into the `%home`. Put a commit message
973−into `home/.commit` and it'll happen. If there's an error, it
978+Want to commit your changes? That means merging the working
979+desk, `%home-bosdut-sigder`, into the `%home`. Put a commit message
980+into `home/.commit` and it'll happen. If there's an error, it
974981will appear as `.error`.
975982
976983The mount has a .version file which simply contains the change
977−number (of the working desk) which any edits are relative to. If
984+number (of the working desk) which any edits are relative to. If
978985the urbit daemon is running and watching, edits will be absorbed
979986quickly (probably after a heuristic timeout), and the version
980−incremented. But we always are confident that the edits are
987+incremented. But we always are confident that the edits are
981988against the latest version, or at least we can use this as a
982−sanity check. All kinds of insane things can happen to get a
989+sanity check. All kinds of insane things can happen to get a
983990filesystem out of consistency.
984991
985992This design obviously needs a little fleshing out, but I'm quite
@@ -988,12 +995,12 @@ from Unix.
988995
989996## In conclusion
990997
991−There's obviously a lot of fun here! There are two things you
998+There's obviously a lot of fun here! There are two things you
992999can do right now.
9931000
9941001One, talk about `%clay` on [this forum
9951002thread](http://urbit.org/~~/fora/posts/~2016.10.25..17.37.47..6937~/) or
996−on [:talk](http://urbit.org/stream). Both hosted on Urbit
1003+on [:talk](http://urbit.org/stream). Both hosted on Urbit
9971004itself!
9981005
9991006Two, poke around the