jamsend - a complication

 Major bit later! On to the stuff the user is going to know.

Alright!

New data layer (Stuff) can produce a UI:Strata about a subset of tree-like stuff:

Then we make those /*%path real:

Ordering is generally not guaranteed by Stuff, so we need %seq. It will probably work fine given the rows are all in one resolve().

Journeying further:

Oh we don't need that,

Yes. Struggling with perfection.

We now give up at, without going into, Disco.

These two container-looking grey lines are the edge of the DirectoryShare and the edge of the DirectoryShares, which are instances of Things/Thing parameterised.

Each have buttons for stop. The one inside is in a set (many Thing items) so has delete, and some controls for manipulating the DirectoryModus.

Recently we have been delving right into the metaphysics going on in the middle of the room defined by a *Modus, we are managing attention on a structure (D**) we can slowly reveal and play with.

D** will also need to vanish as part of its lifecycle, to avoid having too much on its mind...

Here's a wall of noise to give you a feel for developments:

e5b1697 I'm sorry, we're out of time WIP
dbdb958 TimeOffice.*
f5040e2 replace() / n.is() for no history, or A++ won't clear it
3ffad29 launching %A, a train of thought...
d200597 obsolete replacies()
0e7573e tidier lists of actions, C++ restarts Modus memory
93f6fb2 keep ActionButtonson visible with a Scrollability
7a24cef stabilise Strata
0296d75 flat Strata, still over-reative
c29e516 flat Strata acts much nicer
14aab81 some of
0d530d4 target a journey via UI:Strata
f43ad54 to.nice_boost for free-ish %nib:blob
7c5d655 redivide Selection|Directory, spawn Structure
c15610d unstuck via to.match_boost, call it to.distance
5833a33 rearrange, tour %nib:blob gets stuck!
e31cb76 subdividing tour action
c215b90 lose %plodding
d8f772f %journey/(%previously|%forwardsly)
8c66643 move targeting to the %tour
3b50a1a merge ...Utils into Selection
bc23b43 to%tour etc
5c6bac3 default open the top
f17eac0 journey_affects_D in the middle, for depth first
e07d406 move Dierarchy etc to Selection
d9651f1 drop Frontier in favour of the journey
e2bee44 separate transactions on D
b177db2 journey plodding WIP
7d0750e Struggling with perfection
ef34ae6 polish Strata
839a952 configured by %Strata,match|see/*%*
0774871 rename to Strata
f4d118d hanging indent!
d808c90 Filings / Stuffing examines the original would have to clone .X for them to have /*, which might be important
08c2d8b hacky up Filings for now Stuffing a subset
e9bf359 data sanity, top D is %Se,Tree
b6aa035 basic yay-many cursor
afc3abe forward|reverse, stirring around near path deciding
1d340ed make stuffziado?.display_name clickable
2220773 see collections
3774d1e reverse it
24994b3 Stuffziad openness, style
31ebef3 %ads come from %readin=art, accumulate meta
d87a06f defuse pattern of %readin=$k,val=$v, expect more passes
ad8f39c function klepto, %ads hoisting
951e4a3 parse track names, which suggest artist/album/track
6565424 tinkering, C.c|sc typing
6920648 just put traces on %Tree, then we have all indexables in one place, we can assume %name will win?
6c3e8f2 have arguments with Selection//Travel
098f7c8 use for
af87aa1 async tricky
4af922d stashed further than Stuffing...Stuffing seems to require a stashed.version++
e9a9e14 M.stashed, to keep Stuffing** hanging out
522a37c for track names...
8bf7205 Stuff.d() is Travel.dive(), for now
1997aba x2 on first run fixed
1002929 reactivity lacking in Stuffusion x2/*37, use lets_redraw to see
80d0b5f notice the topmost new %Tree, simple renames
4bf18f7 move to Selection.process()
6596b5f hang Travel and its replace()s on n, which replacies() on the Modus
d16854e T.c.y_many
e8171e7 same but different
09fa5eb d() asyncily
a4a7c6a Travel_DLs() misses everything
d442a94 bump
86ce50d give F to any IDBer, schema per F
13200a3 accumulating meta
31e688d d(C,s,T)
00c63df tap things with spanners

As alluded to in the above git log, targeting improved.

We target to improve coherence around the important bits of %Tree. Waking. Creating.

Putting that in a loop is next, for mkdir -p behaviour, then %readin (and %readout?) for reading meaning into the filenames.

I might arrange an audio-thumbnail cache at .jamsend/radiostock/$nowish/$pathhash.webms, which is json encoded %record/%preview, then a \n (escaped in json), then each %preview buffer concatenated.

That's what it's like.

Now we've got a pretty solid... Making stock, having stock...

The green tub above "The Plot Here" is all that's manifested of our connection to some remote music haver or wanter, which hasn't adapted to the new %nib** stuff.

And so, it looks like I can get the Radios system ready for family time...

And then adding another edge comes out quite lovely:

Oh by the way, if your mouse pointer is stuck, unclickable, as a resize window cursor, on xfce, and switching to a console and back doesn't fix it, run: xfwm4 --replace

This grew out of control. I don't think my Select.process() is ever removing anything

Oh yeah! There was a bug in resolve(), which is way in the heart of the thing.

There's also a way to not traverse classing sensibly.

Hmm.. speaking of traverse classing, something's sometimes something.

What about this non-perfect bit...

Yeah. It's almost basically just, now with inhibit(En) etc,

also this little view is always burned in:

It has been getting the debugger compulsively stopping in some silly places.

Setting up production involves a lot of testing and terminality:

and this is different in production:

And so it too much else it seems, so pragmatically, we're just running a development server in production!

But look what that saves:

Good impression of progress. You can see that ramp up to the holidays and then chunk missing.

Whatever. Here's the visual!

Then

Lets not get busted. I finally master the album I (and several others) haven't got out yet, so:

That's chewing on and producing PCM_32LE, unplayable by Strawberry my regular music player.

These go into fre:ac to turn into 24-bit flac while dropping it 5.25db, then I run this command to embed cover art:

find . -maxdepth 1 -name "06*.flac" -print0 | while IFS= read -r -d '' file; do     ffmpeg -i "$file" -i cover.jpg -c:a copy -y -map 0 -map 1 -metadata:s:v title="Album cover" -metadata:s:v comment="Cover (front)" -disposition:v attached_pic "covered/$(basename "$file")"; done

You'll notice that's almost a many-things doer, but it is pointed at one of the two tracks it randomly didn't do.

Then that picture encoding freaks out WebAudio, so I change the cover art with Strawberry, which requires the music is in my collection, then move it back then encode to 100k ogg for the mobile devices.

    

Anyway, then I put a supervisor, instance tyrant, and a public and private music share appserver together...
Then a selenium grid and web UI of remote desktops of them, but the Directory Access API is tricky to bot-nettify like that. It's very easy to close the handle and hard to get it open again - user intervention required. So I threw that away - but it almost works. Chrome needs a parameter somewhere that lowers security for automated test environments.

And... Also, they need to REPL the new language. Oh and we need to really prove C.resolve().
So here we are:

Oh and we need to see it all so:

Or once queuing waves correctly 9efbe97faa05ff1 and having only one of everything though multiple bookmarks reveal it 0ae2566cfe79f31e7523:

About now... I bury the bulk of distractoworld:

But keep a blog somewhere:

And all sorts of:

Is compiling!

Into:

So... Now we go from a hard coded initial document state to...

We have testing and a plurality of tests called Books, already.
We add one. By the way,

%watched:such is a system for delivering an up to date array of C to H.such.
  built on a system for piping C.version++ state changes somewhere
  so you don't have to define an $effect() to use $effect(), if you can bump a C
  UIs (which are also transmit to the top level UI:Otro via %watched) derive from H.such 

There's a blank space because it usually auto-graphs everything in the House being tested (the H%Run)

But the light blue button "noCyto" is on, since the H%Run/LangTiles has an H%Run/Cyto of its own...
And other random details. Have a read.

I'm throwing Peerily under the Story-scope here:

Using a pre-Atime concretion() system declared by %scheme, which you can see in there, and lots of automated testing, I hope to fix all the remaining connection flakiness in the app.

The medium is suggesting to me (read on) that users will pay constant attention to the tab, or open loads of tabs which gradually OhSnap but contain fractions of the program...
Also your identity might be a body of identities, you having only the fraction you need in each location...
Then we can partition off the crashy or dont-want-to-crash (filehandle) parts of the app, via Dexie liveQuery! Future.

loads of tabs: on their always-on computer to provide hosting infra

OhSnap: which is not a state I can recover them from? the chrome extensions to do auto-refresh don't. auto-refresh is alright but you have to keep checking they don't lose their filehandles! the network could report when this happens to you, it might be a fun social game... Being on the radio for someone going into a cave to fix a computer.

The chrome appservers, 3 tabs that love crashing, I tried to make efficiently able to resume from a good state via libvirt! But alas, it needs more something.

Re-Notionalising Earth

jamsend.git 

dbdb958 TimeOffice.*

Thanks.