00:04 |
mats |
sovereigns can act in their own interest |
00:05 |
mats |
https://www.reuters.com/technology/how-crypto-giant-binance-built-ties-russian-fsb-linked-agency-2022-04-22 |
| |
~ 3 hours 10 minutes ~ |
03:16 |
crtdaydreams |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-22#1098025 << will be inclined to use the cl impl. :D might make a few additions that I've been thinking about |
| |
↖ |
03:16 |
dulapbot |
Logged on 2022-04-22 11:06:23 asciilifeform: meanwhile will link to shinohai's alt-vtron. |
03:16 |
crtdaydreams |
would be good opportunity to study some good code |
| |
~ 32 minutes ~ |
03:49 |
signpost |
crtdaydreams: what additions did you have in mind? |
03:50 |
crtdaydreams |
signpost: just an automation of compression and decompression patchsets |
03:51 |
crtdaydreams |
"wrapping" and "unwrapping" vtree tarballs and signing |
03:52 |
signpost |
I could sooner see transparent use of say gzip than tar. |
03:52 |
crtdaydreams |
wai not both tar.gz |
03:53 |
* |
crtdaydreams just used to calling .tar.gz "tarball" |
03:54 |
signpost |
vpatches are to be read by humans. easier to read them if they're individually gzipped rather than wadded together. |
03:55 |
crtdaydreams |
oh. I mean just for downloading them lol |
03:56 |
signpost |
doesn't really make sense to me. |
03:56 |
crtdaydreams |
easier to download one tarball with entire vtree for a proggy and unpack that and read than download a bunch of individual gzipped patches |
03:56 |
crtdaydreams |
plus say i.e. logotron on ascii's site |
03:57 |
crtdaydreams |
I had to manually go through and download each and every one of those patches |
| |
↖ |
03:57 |
signpost |
the ability to sync a remote and local folder already exists, rsync, wget -R, etc |
03:57 |
signpost |
I'd hesitate to try to cram features into the wrong tool. |
03:58 |
crtdaydreams |
hm. you're right. unimportant thought. |
03:58 |
crtdaydreams |
not useful and just bloating. |
03:58 |
signpost |
this kind of thinking is how you got the recursive shitcaking you see in linux already. |
03:59 |
* |
crtdaydreams has not had opportunity to know better until joining #a |
03:59 |
signpost |
compression might be plenty useful; I've got 10s of mb vpatches in pentacle. |
03:59 |
crtdaydreams |
well, if you're making binary blobs then it might not be too big of a deal. |
04:00 |
signpost |
I'm not making binary blobs. |
04:00 |
crtdaydreams |
I might have misinterpreted that thread though, only partially read |
04:00 |
signpost |
the src vpatches in pentacle are all text. |
04:00 |
crtdaydreams |
ok. well ignore then. |
04:00 |
crtdaydreams |
mb |
04:00 |
signpost |
also, knowing better is a discursive process. |
04:01 |
crtdaydreams |
wouldn't that imply accumulating knowledge is therefore non-linear? |
04:02 |
signpost |
no idea man. if you figure out what accumulating knowledge *is*, bravo. |
04:02 |
signpost |
anyway working with v is highly encouraged. |
04:03 |
* |
crtdaydreams seems to have conflated "knowing better" with "knowing more" |
04:04 |
signpost |
when I had a single gigantic shitwad genesis.vpatch for pentacle (which was wrongwrongwrong), I had the thing gzipped. |
04:04 |
signpost |
if such a thing ends up being worthwhile, important to keep the signature pointing at the uncompressed data. |
04:04 |
signpost |
and now having said that, there's a problem, right? |
04:05 |
crtdaydreams |
I getcha. doesn't matter if I work with V or not if I don't actually *make* anything. sort of in a rut where I haven't actually had any form of sustained productivity. anything I'v made ever has been the product of an 8hr sprint at 3AM |
| |
↖ |
04:05 |
signpost |
the attestation of a vpatch signer is to the text, not a particular compression scheme, and yet if you don't have the zipped representation signed you're saying you'll run god knows what from any counterparty through gzip. |
04:05 |
signpost |
(you have this problem too before you verify sig, with gpg) |
04:06 |
signpost |
this might be fine, just showing how you create new problems with additional complexity. |
04:06 |
signpost |
crtdaydreams: the urge to "be productive" is a bad motivator, creates makework. |
| |
↖ |
04:06 |
signpost |
consider instead reading all the blogs that have discussed v, and write about your findings on your own site. |
04:06 |
signpost |
perhaps then you do end up with something worthing fixing or adding. |
04:07 |
signpost |
I've been working on pentacle off and on for over a year now, and have not released it because it's not yet worth anyone else's time or effort. |
04:08 |
signpost |
this doesn't bother me; the work on it has more to do with my own needs (in re: server management, but also other things) than wanting to be seen as productive. |
04:08 |
crtdaydreams |
maybe something lost in transmission but I was under the impression the compressed tarball would have sigs and patches in it? |
04:08 |
crtdaydreams |
not just patches |
04:09 |
signpost |
the reason I think that idea should be dismissed is that "download all the related things" is a distinct problem |
04:09 |
signpost |
(may soon be done over pest by entirely other means, for example) |
04:09 |
signpost |
then I was discussing the thought of individual gzipping of vpatches, and gave some reasons it's complicated. |
04:10 |
signpost |
maybe it's complicated but useful, and gets done. these aren't good-and-evil evaluations. |
04:11 |
crtdaydreams |
^ speaking of pest re: DHT I was thinking of a 1-byte principle similar to the vpatch one; e.g. detailing 11 "have read and understand". similar format could be used for "availiable on request" "availiable to only users in L1" etc. |
04:11 |
signpost |
could you elaborate; I don't get it. |
04:14 |
crtdaydreams |
trying to find the article, I think it was on trilema, but it was a single-byte representation at the start of a vpatch. |
04:15 |
crtdaydreams |
and detailed i.e. "I have read and understand the whole." "I wrote majority of it." etc. |
04:19 |
crtdaydreams |
signpost: found it |
04:22 |
crtdaydreams |
I think it would be useful in a DHT extension as you could specify in 1-byte who can request what |
04:23 |
crtdaydreams |
i.e. for a file in a /.dht/ dir. if listed as 0101 might be available "only upon direct request for $hash" "by $user in L1" |
| |
↖ |
04:25 |
crtdaydreams |
whearas another file might be 1111 "publicly advertised $hash" "to all $peer (s)" |
04:26 |
crtdaydreams |
or " |
04:26 |
crtdaydreams |
or "to $peer of $peer" etc. |
04:29 |
crtdaydreams |
would be willing to write up a full single-byte vector for DHT |
04:41 |
signpost |
in what case would I want to admit to existence of a file to a counterparty that may not retrieve it? |
04:45 |
verisimilitude |
The technical term is ``taunting'', signpost. |
04:45 |
signpost |
lol |
04:46 |
signpost |
fucking guy gave me the neener bit! |
| |
~ 39 minutes ~ |
05:25 |
crtdaydreams |
signpost: peer forwarding |
05:25 |
crtdaydreams |
but ``taunt'' for lolz could also be pretty funny |
| |
~ 42 minutes ~ |
06:08 |
crtdaydreams |
signpost: that's the thing. having another bit that determines availability is another thing again. can set e.g. 0111 "only upon direct request for $hash" "to $peer of $peer" |
06:08 |
crtdaydreams |
so anyone who has the hash and is a friend-of-a-friend could ask for that file. |
06:08 |
crtdaydreams |
peer forwarding would have to be optional ofc |
06:10 |
crtdaydreams |
useful for files like e.g. source to be available to anyone, websites, etc. |
06:12 |
crtdaydreams |
immediate $peer that is forwarding could restrict the availability of a file on their station, so on $host it might be 1111, but on $peer forwarding, it might be 0101 |
06:14 |
crtdaydreams |
can be argued that "if $peer2 of $peer1 wants a file, then $peer1 ought to download it, then share it with $peer2 |
| |
↖ |
06:17 |
crtdaydreams |
which probably fits more inline with pest philosophy, but I still think having the capacity to make information freely available should have precedence, at least an option for forwarding that is clearly defined in spec |
| |
↖ |
| |
~ 7 hours 11 minutes ~ |
13:28 |
mats |
https://archive.ph/vDBR7 |
| |
↖ |
13:41 |
shinohai |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-22#1098232 << Did you ever manage to stand up a blog or something? Would love to read more if you decide to work on this. |
| |
↖ |
13:41 |
dulapbot |
Logged on 2022-04-22 23:14:53 crtdaydreams: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-22#1098025 << will be inclined to use the cl impl. :D might make a few additions that I've been thinking about |
13:43 |
shinohai |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-22#1098246 << Laziness on my part had me put vtrees into subdirectories on my www, and I simply made a bash function (w/ wget as signpost mentions) so I can gentoo style `sync <vtree>` to populate it locally. |
13:43 |
dulapbot |
Logged on 2022-04-22 23:56:04 crtdaydreams: I had to manually go through and download each and every one of those patches |
| |
~ 1 hours 30 minutes ~ |
15:14 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098289 << this is rather redundant in pestronics -- recall that a pest station communicates ~exclusively~ with its l1 peers. |
15:14 |
dulapbot |
Logged on 2022-04-23 00:22:05 crtdaydreams: i.e. for a file in a /.dht/ dir. if listed as 0101 might be available "only upon direct request for $hash" "by $user in L1" |
15:15 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098305 << exactly |
15:15 |
dulapbot |
Logged on 2022-04-23 02:13:20 crtdaydreams: can be argued that "if $peer2 of $peer1 wants a file, then $peer1 ought to download it, then share it with $peer2 |
15:17 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098306 << you can mark this, but the technical term for such a mark is 'promisetronic', i.e. you can't force anybody to obey the mark |
15:17 |
dulapbot |
Logged on 2022-04-23 02:15:47 crtdaydreams: which probably fits more inline with pest philosophy, but I still think having the capacity to make information freely available should have precedence, at least an option for forwarding that is clearly defined in spec |
15:18 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098307 << will be lulzy when Official financialism becomes vulnerable to ethertard-style 'dao' bugolade |
15:18 |
dulapbot |
Logged on 2022-04-23 09:27:10 mats: https://archive.ph/vDBR7 |
15:18 |
mats |
inshallah |
15:20 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098272 << recall the lulz club where diana et al, 'let's write reports about plan to write reports!' etc. |
15:20 |
dulapbot |
Logged on 2022-04-23 00:05:09 signpost: crtdaydreams: the urge to "be productive" is a bad motivator, creates makework. |
15:21 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098268 << exactly this. |
15:21 |
dulapbot |
Logged on 2022-04-23 00:03:48 crtdaydreams: I getcha. doesn't matter if I work with V or not if I don't actually *make* anything. sort of in a rut where I haven't actually had any form of sustained productivity. anything I'v made ever has been the product of an 8hr sprint at 3AM |
| |
~ 15 minutes ~ |
15:36 |
billymg |
morning asciilifeform |
15:37 |
billymg |
i'm working on the backlinks feature for the logotron, noticed in the schema you already have a backlinks field of type integer[] |
15:38 |
billymg |
that would only work for backlinks within a channel though, unless i'm missing something |
15:39 |
billymg |
to support cross-channel backlinks would also need to specify the channel |
15:42 |
shinohai |
In other ethtard news: https://archive.ph/MmDV2 |
| |
~ 53 minutes ~ |
16:35 |
asciilifeform |
wb billymg . indeed had a placeholder for backlinks, but never implemented the meat of it at all |
16:36 |
asciilifeform |
billymg: asciilifeform's orig. notion was to simply use row index in that field, so not need to specify channel, would point to all db row(s) where linked the current row |
16:37 |
billymg |
asciilifeform: ah, ok that makes sense then |
16:39 |
asciilifeform |
schema, seen via phf's tool for ref. |
16:40 |
billymg |
yeah, i'll use 'ser', that works |
16:41 |
asciilifeform |
billymg: there's a caveat, ser may not presently work, as it aint the 'primary key' |
16:41 |
* |
asciilifeform not even tested |
16:42 |
asciilifeform |
it also aint exported in the raw-log view, also |
16:43 |
asciilifeform |
(and neither is the backlink field for that matter) |
16:44 |
* |
asciilifeform suspects that backlinkage really gotta be calculated locally to work consistently |
16:44 |
phf |
i've always treated backlinks as ephemera (and other kinds of `annotations`) |
16:44 |
asciilifeform |
phf: right, imho has to |
16:45 |
billymg |
phf: did you store backlink ids in the db anywhere? |
16:45 |
phf |
lol "db" |
16:45 |
billymg |
or on disk |
16:45 |
asciilifeform |
billymg: he had whole thing in memory, rather than dbism |
16:45 |
billymg |
oh lol |
16:45 |
asciilifeform |
thread |
16:45 |
dulapbot |
Logged on 2022-04-07 22:55:24 phf: asciilifeform: original cmucl version was mmaped, but when porting to sbcl due to frequent crashes had to make the whole thing in memory, which is also major reason why ia haven't expanded it. needs to be made mmaped again, to support multiple channels. otherwise unreasonably memory heavy |
16:46 |
billymg |
phf: could you expand on how you handled backlinks? |
16:47 |
billymg |
asciilifeform: re: 'ser', why does it matter if not the primary key as long as it's globally unique? |
16:48 |
* |
billymg tested and SELECT COUNT(DISTINCT(ser)) FROM loglines; == SELECT COUNT(ser) FROM loglines; |
16:49 |
asciilifeform |
hm asciilifeform distinctly recalled that later found in the docs 'gotta set it as primary key if want guaranteed unique' but nao cannot find any such thing |
16:49 |
phf |
billymg: it's pretty dumb, and by dumb i mean straightforward. i have annotation mechanism that looks at log line and thinks what else to say about it, so e.g. if there's a link there, it just goes to the original message and plops an extra annotation on it. this is all extremely cheap and easy because working with structs and arrays |
16:50 |
asciilifeform |
phf: iirc you also had 'bot' and 'reader' integrated in 1 proggy? this'd also make above easier, would expect |
16:50 |
phf |
yes |
16:52 |
billymg |
phf: how did you store the logs on disk? |
16:52 |
billymg |
and were the annotations included? |
16:52 |
* |
asciilifeform used sqlism in his logotron mainly so as to separate 'bot' and 'reader'. which is, as phf illustrates, double-edged sword, makes certain things harder |
16:53 |
phf |
billymg: logs are store in kako format, and i don't include annotations. they are treated as ephemera. things are reannotated during load |
16:53 |
billymg |
kako? |
16:54 |
asciilifeform |
billymg: the format seen in raw-log |
16:54 |
dulapbot |
Logged on 2022-03-27 11:13:43 asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2022-03-27#1089440 << there's a raw-log knob, which yields up to 500 ln, e.g. |
16:54 |
phf |
does he predate you? kakobrekla, the original log keeper amongst other things |
16:54 |
* |
asciilifeform refs to it as 'phf format' but dates to era1 |
16:54 |
billymg |
ah, yeah, i joined post kako |
16:55 |
billymg |
in the context i thought maybe some data format |
16:56 |
phf |
well, yeah :) it's in my mind the canonical log format. technically my logs are only stored in kako format, everything else is runtime side effects |
16:56 |
asciilifeform |
exactly like this, but also includes channel |
16:56 |
asciilifeform |
or hm, possib. original didn't include it |
16:57 |
asciilifeform |
ah yea, asciilifeform's importer does ask for channel and 'era' |
16:58 |
asciilifeform |
subj. |
16:58 |
billymg |
asciilifeform: ah, i see what you're saying |
16:58 |
billymg |
'ser' will only work if generated by the operator, won't be "portable" (because different operators can have different list of chans) |
16:58 |
billymg |
that's the first part that i'm working on now anyway, the "backfill" script |
16:59 |
asciilifeform |
billymg: so long as you regen the backlinks after any raw db manipulations , oughta work |
17:00 |
billymg |
asciilifeform: alternative is a chan/idx compound key |
17:00 |
asciilifeform |
( in particular , any db mutilations which remove lines ) |
17:00 |
billymg |
makes it more portable |
17:00 |
asciilifeform |
that'd also work, if you don't do anyffin exotic like swallowing sumbody else's db into existing |
17:00 |
phf |
billymg: yeah, i'm sorry, the btcbase architecture is totally custom, doesn't map to "lamp" type stack |
17:02 |
* |
asciilifeform regards phf's method as smarter, keeping whole thing in ram is in fact over9000 cheaper than postgresism + its caches |
17:02 |
billymg |
phf: it was helpful for understanding another way of doing it, and what the tradeoff is with the db model |
17:03 |
asciilifeform |
in particular, making search go in realtime in asciilifeform's logotron requires rather gnarly and postgres-specific indices mechanism, and still not anywhere as snappy as phf's search, where simply whole thing in ram |
17:05 |
billymg |
asciilifeform: using \d+ in postgres shows the complete loglines table at 300MB |
17:05 |
billymg |
so could even store in memory on the RK |
17:05 |
asciilifeform |
easily |
17:06 |
asciilifeform |
if one were to represent links compactly (e.g. separate field where stores offsets into text where they belong) could be evenmoar tight |
17:06 |
asciilifeform |
*log links |
17:07 |
phf |
billymg: 300mb? |
17:07 |
phf |
oh that's for #asciilifeform |
17:07 |
phf |
how large is #trilema? |
17:07 |
asciilifeform |
a rather embarrassing % of the text consists of things like http://logs.nosuchlabs.com/log/asciilifeform |
17:08 |
billymg |
phf: that's for all the chans in my db |
17:08 |
billymg |
trilema, asciilifeform, pest, ossasepia, etc. etc. |
17:08 |
phf |
weird, hmm, i wonder what they do. is it stored compressed smh |
17:08 |
billymg |
ah, yeah, not sure how pg is calculating that |
17:09 |
billymg |
could be compressed |
17:09 |
billymg |
though asciilifeform's gzipped dump totals ~80mb, so 300mb seems right |
17:09 |
asciilifeform |
this hour's backup is ~82MB gz'd, ~266 -- unzipped |
17:10 |
asciilifeform |
( this represents all chans that live in asciilifeform's logotron ) |
17:10 |
asciilifeform |
the indices (which do not appear in the backup) of course weigh sumthing, tho |
17:10 |
asciilifeform |
and occupy memory/disk |
17:11 |
billymg |
asciilifeform: the table + indices is almost 900mb (897 according to \l+ in pg) |
17:11 |
phf |
hmm |
17:11 |
asciilifeform |
billymg: sounds right |
17:12 |
asciilifeform |
this, plus the (typically machine-specific, iirc mine's set to 4GB!) cache, makes the thing a distinct hog |
17:12 |
dulapbot |
Logged on 2022-04-07 22:58:54 asciilifeform: e.g. asciilifeform has iirc 4g cache set just for postgres. so stood next to ~that~, light.. |
17:12 |
phf |
there's some modifications of kmp over lz77 and such. maybe if i ever pick up btcbase code again.. |
17:13 |
asciilifeform |
phf: asciilifeform suspects that compressing individual lines aint +ev. but cannot easily prove |
17:13 |
phf |
asciilifeform: shared dictionary |
17:13 |
asciilifeform |
for 1 thing, gotta preserve searchability |
17:14 |
asciilifeform |
phf: a la verisimilitude, heh. possibly then. |
17:14 |
phf |
kmp in above is knuth-moris-prat |
17:14 |
phf |
possibly searchability is the only confounding factor |
17:14 |
asciilifeform |
imho the data set dun weigh enuff (and unlikely to , any time soon) to justify this kinda trade |
17:15 |
phf |
asciilifeform: ironically i spent soem time trying to explain to verisimilitude that his idea is not novel, and lz77 e.g. had pre-computed/shared/etc. dictionaries since forever. henceforth and therefore it shall go in the logs as "verisimilitude idea" |
| |
↖ |
17:16 |
* |
asciilifeform recalls |
17:18 |
* |
asciilifeform previously tried to explain to him that maintaining shared data sets aint at all 'phree', and the cost is substantial and paid up-front, whereas the 'win' is imho tenuous. but can't be arsed to find in log atm |
| |
↖ ↖ |
17:19 |
asciilifeform |
you'd need some scenario no less exotic than e.g. an interplanetary pestnet, at 'dollar a byte' bandwidth, to justify such thing imho |
17:21 |
asciilifeform |
( and even there -- open question whether +ev ) |
17:21 |
* |
asciilifeform must bbl |
| |
~ 1 hours 1 minutes ~ |
18:23 |
verisimilitude |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098416 Elision is gestalt, phf. |
18:23 |
dulapbot |
Logged on 2022-04-23 13:14:13 phf: asciilifeform: ironically i spent soem time trying to explain to verisimilitude that his idea is not novel, and lz77 e.g. had pre-computed/shared/etc. dictionaries since forever. henceforth and therefore it shall go in the logs as "verisimilitude idea" |
18:24 |
verisimilitude |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098418 Returning to the Latin, consider storing all ancient Latin with but one dictionary. The advantages may only come with such sharing, or may come before; I don't yet know. |
18:24 |
dulapbot |
Logged on 2022-04-23 13:17:21 asciilifeform: previously tried to explain to him that maintaining shared data sets aint at all 'phree', and the cost is substantial and paid up-front, whereas the 'win' is imho tenuous. but can't be arsed to find in log atm |
18:25 |
verisimilitude |
Do I need to explain how something only useful to a decompression algorithm is distinct from Elision, phf? |
18:31 |
verisimilitude |
I've already explained how the representation is useful for operations other than being expanded into characters. |
| |
↖ |
| |
~ 2 hours 21 minutes ~ |
20:52 |
adlai |
your use of this loanword is... far from how I understand it. |
20:53 |
adlai |
verisimilitude: what does 'gestalt' mean, preferably without straying far outta the one thousand words of simplified english? |
20:53 |
* |
adlai is not kidding |
20:55 |
adlai |
in other 'not kiddings' -- I doubt I've read a more confusing analogy in my life, than conflating programs and programmers |
20:55 |
dulapbot |
Logged on 2022-04-22 12:57:08 asciilifeform: pictures cmucl as sumthing like 'trb' in re sbcl's 'prb' |
20:56 |
adlai |
this is a particularly bad combobulation of similarities, because SBCL was indeed some sorta 'hostile takeover' (to borrow the capitalist definition of hostility, without bloodshed! only feelings hurt!); |
| |
↖ |
20:57 |
adlai |
however, if anything, the process begun when asciilifeform started grinding geneses of trb, rotor, stator, etc, is actually quite similar to what Newman did to the CMUCL human-guided self-patching binary. |
20:58 |
adlai |
so... quite an antipersonnel mine of divergent similarities |
21:00 |
adlai |
back to the most recent stack frame of others: my understanding of the human linguistics affecting verisimilitude's project is that, if you're trying to learn a foreign language, the compression dictionary of its spellings is worse than unhelpful; however, ... |
21:02 |
adlai |
... it could be that classical Latin is a language from early enough in history (i.e., written history, as opposed to hearsay and fossils) that it behaves differently from mein freund hat mir ein hoch gift gegeben, wunderbar! |
21:04 |
adlai |
[ for those who haven't studied mistranslations between English and German: 'mein' until 'wunderbar' is a phrase containing the word 'gift', which means poison in German ] |
21:06 |
adlai |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098418 << whenever I find myself thinking of ideas similar to verisimilitude's Elision project, I wonder about the cost of generating such datasets, rather than how heavy the generated dictionary would be. |
21:06 |
dulapbot |
Logged on 2022-04-23 13:17:21 asciilifeform: previously tried to explain to him that maintaining shared data sets aint at all 'phree', and the cost is substantial and paid up-front, whereas the 'win' is imho tenuous. but can't be arsed to find in log atm |
21:07 |
phf |
adlai: first one should establish the difference between futurism and engineering |
| |
↖ |
21:07 |
adlai |
'cost' might be the wrong word for the speculative process ending in "ain't nobody got time nor need for doing that" |
21:08 |
adlai |
if those are the two pursuits being disambiguated, I'd call engineering almost equivalent to 'future-proofing', whereas futurism is more along the lines of "how dare you ask why I collect art.... WHAT IS THE USE OF A BAAAABBBBYYY?" |
21:08 |
phf |
if one's doing futurism primarily, then costs exists are equivalent to other considerations, that is they are a talking point, rather than a practical consideration. "cost will be 0, because the interplanetary commission will fund the implementation, on the grounds that it is of the ultimate importance to human kind!" |
21:12 |
adlai |
e.g., if I decide to put into practice the Elision idea -- as described in the logs, since I've not yet grokked the moving target in verisimilitude's publications -- except, I stop caring about Latin and want to do this for Aramaic texts because that's what was the business language of Jesus's Palestine, then the 'costs' of generating the dictionary would be obtaining and processing the |
21:12 |
adlai |
entire corpus, in addition to the text which I wish to in/deflate. |
21:14 |
adlai |
currently I give benefit of reasonable doubt to the possibility that there might be an abstract method of generating such a dictionary, instead of simply walking every single document in the relevant language from an arbitrary window of time. |
| |
↖ |
21:15 |
adlai |
phf: do you cut between these elsewhere than I did? |
21:18 |
* |
adlai considers the fact that folks work off different, wildly different definitions as one of the great conversational evils; e.g., ... |
21:20 |
* |
adlai when encountering the word 'gestalt', for example, first dereferences to "linear versus gestalt", as described in Greg Egan's novel Diaspora; and only subsequently to "blech, maybe if I look it up in wiktionary, it'll tell me that I never understood the correct hochdeutsch definition either..." |
21:21 |
adlai |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098427 << where? |
| |
↖ |
21:21 |
dulapbot |
Logged on 2022-04-23 14:29:53 verisimilitude: I've already explained how the representation is useful for operations other than being expanded into characters. |
21:24 |
verisimilitude |
To be gestalt is to be more than the sum of parts. |
21:25 |
verisimilitude |
http://logs.nosuchlabs.com/log/asciilifeform/2022-02-25#1080748 Try to solve this well, adlai. |
21:25 |
dulapbot |
Logged on 2022-02-25 13:26:38 verisimilitude: Consider the problem of searching from one point to another, forwards or backwards, for the first word beginning with some specified letters. Let me know when to share my solution. |
21:26 |
adlai |
if you give any partial description of your solution, then I'll have some idea of what 'well' means, other than 'better than good enough'. |
21:27 |
verisimilitude |
http://logs.nosuchlabs.com/log/asciilifeform/2022-02-25#1080884 |
21:27 |
dulapbot |
Logged on 2022-02-25 22:16:41 verisimilitude: Get the index of the first possible such word, and of the last possible such word; now the searching is a bounds check on each word, needing no more dictionary access. Searching the auxiliary dictionary would mean just an additional bounds check. Isn't this neat? Certainly, turning a letter problem into an integer problem satisfies this. Thi |
21:27 |
verisimilitude |
I should've chosen a different message to quote. |
21:28 |
adlai |
e.g., in your format, the entire document has already been processed; if I land in the middle of an unseen document, then do I first need to ask for a map of the territory, before I can begin searching? |
21:29 |
verisimilitude |
Nevermind. |
21:29 |
verisimilitude |
Yes, I suppose so. |
21:29 |
* |
adlai was thinking more along the lines of "the naive solution is polynomial in the size of this, exponential in the size of that" |
21:29 |
verisimilitude |
The same holds true for ASCII and Unicode, note. |
21:30 |
verisimilitude |
No, adlai, I want constant and linear time algorithms as much as possible. |
21:30 |
adlai |
right. that's a much more concrete critereon than "solve this well"! |
21:31 |
verisimilitude |
Here's a different phrasing of my solution. |
21:31 |
crtdaydreams |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098308 << you've inspired a sprint to the finish |
21:31 |
dulapbot |
Logged on 2022-04-23 09:40:09 shinohai: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-22#1098232 << Did you ever manage to stand up a blog or something? Would love to read more if you decide to work on this. |
21:31 |
verisimilitude |
Get the index of the first possible such word, and of the last possible such word; now the searching is a bounds check on each word, needing no more dictionary access. Searching the auxiliary dictionary would mean just an additional bounds check. Isn't this neat? |
21:31 |
verisimilitude |
This works because the dictionary is sorted, and even if it weren't, there could be a transformation table to make it so; thus, other problems can have such tables and use such simple methods. |
21:32 |
verisimilitude |
I want to turn problems of language into problems of numerical figuring, adlai. |
21:32 |
adlai |
you want to make paper vanish, and deal only with engraved stones. |
21:33 |
verisimilitude |
I want to make something better than paper, but I may never be able to do so. |
21:33 |
crtdaydreams |
also thank you adlai for showing me coleslaw. in last ~20 min setting everything up it's been quite a pleasant tool to work with, despite having googleware bloat and shithubism built-in. |
21:34 |
verisimilitude |
http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098447 What's wrong with this method? |
21:34 |
dulapbot |
Logged on 2022-04-23 17:12:47 adlai: currently I give benefit of reasonable doubt to the possibility that there might be an abstract method of generating such a dictionary, instead of simply walking every single document in the relevant language from an arbitrary window of time. |
21:34 |
verisimilitude |
I also enjoy coleslaw. |
21:34 |
* |
crtdaydreams for the life of him could not get fastcgi working and ditched it. best move could make than farting about with phpism |
21:34 |
adlai |
crtdaydreams: you don't have to serve any of that in the generated pages, no more than you need to compile all of CLOS into javascript because you want a gif generated client-side! |
21:35 |
crtdaydreams |
adlai: hehehe, I've already disabled the lot of it |
21:35 |
verisimilitude |
Ideally, someone else has already made a suitable dictionary. |
21:37 |
adlai |
I find the connotations of calcification quite chilling, that's all. maybe I'm giving too much credit to the joys of being alive, and oughtta go back to speaking only well, and only of death. |
21:37 |
verisimilitude |
There's an auxiliary dictionary for just this, adlai, and it must be the second system added to the floor of Elision. |
21:37 |
* |
adlai does recall, fwiw, mini-sermon in Latin class about "do not speak ill of the dead, because only you are present to defend them" |
| |
↖ |
21:38 |
verisimilitude |
I don't want to restrict language, but I do want something more suited to storing proper language. |
21:38 |
adlai |
if you've looked over parts of the logs from the early days ... how frequently do you think would be reasonable to regenerate auxiliary dictionaries? |
21:38 |
verisimilitude |
Storing and manipulating, to clarify. |
21:39 |
verisimilitude |
I don't. |
21:39 |
adlai |
one advantage of asking this in the context of IRC logs is that there are easy foundations; e.g., once per calendar day, stfu; once per change of ratings graph; etc |
21:40 |
verisimilitude |
Ask a designer of vector graphics how he would store a photograph of subway graffiti, adlai. |
| |
↖ |
21:40 |
adlai |
bwahahaha what do you think I do when I stand arm's length away from the wall? |
21:40 |
adlai |
hint: I'm not reading the letters. |
21:41 |
* |
adlai imagines some folks see him standing there, and wonder -- has the idiot learned too many alphabets? |
21:41 |
verisimilitude |
Has anyone here ever created a new symbol for the conversation, or only used ASCII and Unicode? |
21:42 |
verisimilitude |
Not only do character sets restrict us, adlai, they don't serve a good job, as I aim to prove. |
21:43 |
verisimilitude |
They are middling, and I want extremes. |
21:43 |
adlai |
I've destroyed drafts of glyphs for numerals, mainly because one of my goals is for the fundamental properties of the shapes to be sufficiently general that I could re-derive them. |
21:43 |
* |
adlai has never attempted doing this for phonetic letters |
21:44 |
adlai |
these days, that project is on indefinite hiatus, because I've been focusing on producing prose again, using conventional keyboards. oh well. |
21:44 |
verisimilitude |
Some of mine are here. |
21:44 |
adlai |
why do you have no named links? |
21:44 |
verisimilitude |
I may want to change the names. |
21:45 |
verisimilitude |
I've been designing a base-thirty glyph set. |
21:45 |
adlai |
when you edit an article enough to rename it, wouldn't that be enough of an act for much time to have elapsed, anyway? |
21:46 |
adlai |
fwiw, I've not commented on that article yet, mainly because upon reading it, I realized that our goals are completely orthogonal to eachother. |
21:46 |
verisimilitude |
Perhaps. |
21:47 |
verisimilitude |
Once I finish, the number symbols will be moved to a new article. |
21:47 |
adlai |
maybe even slightly opposed; I don't enjoy illiteracy, despite the fact that I frequently wish to return to such a state. |
21:47 |
verisimilitude |
Look at these pretty things, adlai. |
| |
↖ |
21:48 |
adlai |
are these still the glyphs that work for calculator-style LED displays? |
21:49 |
verisimilitude |
Which? |
21:50 |
adlai |
these, those, the "pretty things" in the articles you link. |
21:51 |
verisimilitude |
No. |
21:51 |
verisimilitude |
A glance will show as much. |
21:52 |
verisimilitude |
Oh. |
21:53 |
verisimilitude |
It can't be helped. |
21:53 |
* |
adlai|textonly has not yet gone blind! only, stubborn. |
21:54 |
verisimilitude |
They've not yet been rendered to raster graphics, and exist only as SVG for now. |
21:55 |
verisimilitude |
Many SVG renderers ignore the line command of all things. |
21:56 |
adlai|textonly |
have you compared APL with the various recent linenoiselangs inspired by it? e.g. 'j' |
21:56 |
* |
adlai|textonly mostly encountered these in solutions posted to projecteuler.net problems |
22:00 |
verisimilitude |
I dislike J. |
22:00 |
verisimilitude |
I dislike K also. |
22:00 |
verisimilitude |
The glyphs are most valuable, and I won't do without them. |
22:01 |
adlai|textonly |
how does GNU APL cope with terminals? |
22:01 |
verisimilitude |
I may acquiesce and write a convertor for myself, at some point, in order to use them. |
22:01 |
verisimilitude |
APL existed before the blight, and so was grandfathered into Unicode. |
| |
↖ |
22:02 |
* |
adlai|textonly has considered dedicating whatever future effort to configuring terminal fonts including a handful of unicode codepages, instead of drinking the graphical window system bathwater again. |
22:02 |
verisimilitude |
I must be elsewhere now, adlai|textonly, but I'll return. |
22:03 |
adlai|textonly |
safe travails! |
22:05 |
* |
adlai|textonly meanwhile located http://logs.nosuchlabs.com/log/asciilifeform/2021-05-03#1035390 as the "I'm Feeling Likely" for http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098451 |
22:05 |
dulapbot |
Logged on 2021-05-03 16:11:47 verisimilitude: I discuss this and more in the recently rewritten article. |
22:05 |
dulapbot |
Logged on 2022-04-23 17:20:09 adlai: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-23#1098427 << where? |
22:12 |
adlai|textonly |
you might be able to improve sparsity of your word table's stored representation by eliding all short words computable as linear functions of longer words similar to each other. |
| |
↖ |
22:13 |
* |
adlai|textonly prefers not to call this a 'dictionary'; those typically have pronounciation advice! |
22:13 |
adlai|textonly |
if I'm on mudslinging, fwiw, "Elision" means somewhat the opposite of what you're doing. |
22:14 |
adlai|textonly |
I guess I should have written, 'elision' ; i.e., the actual improper noun means pretty much the opposite of what you're doing with your proper-noun-named project. |
22:15 |
adlai|textonly |
of course, it's your project, you can name it whatever you want. |
22:17 |
adlai|textonly |
the suggestion for reducing sparsity is probably contrary to your preference for extreme solutions. |
| |
~ 1 hours 21 minutes ~ |
23:38 |
adlai|textonly |
one plausible usecase for this kind of agonised linguistic ossification is court records; yer honner, the technical expert's jargon requires a motion to adjourn so the stenographer can recompute auxiliary wordlists |
23:38 |
billymg |
asciilifeform: in looking closely at these now, straightforward regex conversion might not work, the indexes are off |
| |
↖ |
23:38 |
bitbot |
Logged on 2021-07-12 16:30:36 punkman: asciilifeform: these log links are broken in logs "http://logs.ossasepia.com/log/trilema/2020-03-12#1234567" |
23:38 |
bitbot |
Logged on 2021-07-12 16:30:55 asciilifeform: punkman: oh ha i'ma have to add these to the regexp which substs |
23:39 |
billymg |
same with logs.ericbenevides links |
23:39 |
adlai|textonly |
much simpler to just repeat yourself instead of pretending like hyperlinks work. |
23:39 |
shinohai |
!!v FB4DD4A693A4F0791F6E3E8152CF7E4583B88222E2342BE381A3CD6C3E2A08B1 |
| |
↖ |
23:39 |
deedbot |
shinohai rated crtdaydreams 1 << Lisp, Writes at: http://0xcdd.com/ |
| |
~ 18 minutes ~ |
23:58 |
billymg |
actually, possibly the ossasepia links are fine (i thought i hit one that wasn't but now a few more spot checks are resolving fine). a few more random checks with ericbenevides are only resulting in misses |