signpost meanwhile upgraded to gnat-2017, because this didn't require any patching to target musl. removed a shitload of nasty patch cruft I'd accumulated tilting at the gnat-2016 windmill.
verisimilitude: Firstly, imagine having a one megabyte hard disk drive, signpost.
phf: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-20#1097715 << if you're going to mess with fhs, then there's no reason to keep /usr at all
dulapbot: Logged on 2022-04-20 23:53:40 signpost: anyone have a good argument for both /{bin,sbin} and /usr/{bin,sbin} existing?
phf: traditional lore is that "/usr/ is there because / on the canonical unix pdp ran out of space once"
phf: at some point /usr/ separation was justified by special permissions on fs level, e.g. on freebsd in the early 2000s it was suggested to mount / rw, and /usr r (since it's all static and compiled stuff)
phf: here's what e.g. stali does, https://sta.li/filesystem/
phf: "/usr -> / # softlink pointing to /"
phf: oh right, https://twitter.com/foone/status/1059310938354987008, /usr used to be user directories, which is how they have it on plan9 still
shinohai: lmao from the sta.li page "/sucks - stuff that sucks, like ugly gnu library dependencies, or systemd fake handlers"
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097725 << iirc exactly so
dulapbot: Logged on 2022-04-21 09:23:27 phf: traditional lore is that "/usr/ is there because / on the canonical unix pdp ran out of space once"
thimbronion: Mark Zuckerburg banned from Russia in perpetuity for "creating a russophobic agenda"
asciilifeform: lol
asciilifeform: thimbronion: arsebook already banned in ru last month as 'extremist org'
asciilifeform: $ticker btc usd
busybot: Current BTC price in USD: $42286.47
asciilifeform: !w poll
watchglass: Polling 15 nodes...
watchglass: 205.134.172.27:8333 : Alive: (0.023s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883 (Operator: asciilifeform)
watchglass: 205.134.172.26:8333 : Alive: (0.081s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=732883
watchglass: 205.134.172.4:8333 : (172-4.core.ai.net) Alive: (0.143s) V=70001 (/therealbitcoin.org:0.7.0.1/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 205.134.172.6:8333 : (172-6.core.ai.net) Alive: (0.081s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=732883
watchglass: 71.191.220.241:8333 : (pool-71-191-220-241.washdc.fios.verizon.net) Alive: (0.154s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883 (Operator: asciilifeform)
watchglass: 205.134.172.28:8333 : Alive: (0.022s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=732883 (Operator: whaack)
watchglass: 54.39.156.171:8333 : (ns562940.ip-54-39-156.net) Alive: (0.179s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 208.94.240.42:8333 : Alive: (0.189s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 54.38.94.63:8333 : (ns3140226.ip-54-38-94.eu) Alive: (0.325s) V=88888 (/therealbitcoin.org:0.8.88.88/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 94.176.238.102:8333 : (2ppf.s.time4vps.cloud) Alive: (0.307s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732636
watchglass: 82.79.58.192:8333 : (static-82-79-58-192.rdsnet.ro) Alive: (0.336s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 103.36.92.112:8333 : (terebe.ns01.net) Alive: (0.601s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 103.6.212.28:8333 : Alive: (0.578s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=480191 (Operator: whaack)
watchglass: 75.106.222.93:8333 : Alive: (0.566s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=732883
watchglass: 143.202.160.10:8333 : Busy? (No answer in 100 sec.)
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097722 << 1st and smallest hdd asciilifeform ever owned was 20MB
dulapbot: Logged on 2022-04-21 03:51:00 verisimilitude: Firstly, imagine having a one megabyte hard disk drive, signpost.
asciilifeform: (pc xt)
asciilifeform: prior to that, had 'iskra' w/out hdd, then c64, ditto
asciilifeform: can't recall ever seeing a 1MB hdd.
asciilifeform had built linuxen that sat down in 4MB eeproms, however.
signpost: yeah, was asking whether a good argument remains to retain it.
signpost: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097723 << did in fact remove it already, was running this decision by others.
dulapbot: Logged on 2022-04-21 09:22:40 phf: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-20#1097715 << if you're going to mess with fhs, then there's no reason to keep /usr at all
signpost: this results in the following root structure http://paste.deedbot.org/?id=jADU
signpost: another option would be to put everything that regards programs in /usr, which would leave http://paste.deedbot.org/?id=zMuD
signpost: nice thing is that these are adjustable by the operator of pentacle. each "pbuild" respects PREFIX and DESTDIR.
signpost: this falls out of the tool being designed as a machine that stamps down a userland wherever you may want one, and not strictly a "distro installer" with strong opinions about what installation means.
signpost: for that matter, it doesn't even have an opinion about what software comprises a given userland. depends on what vtrees you have, and which you choose to build.
signpost: (aside that the one in which builds occur has the obvious dependencies, gnat, vtools, musl, posix shell, etc)
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097767 << imho this is The Right Thing
dulapbot: Logged on 2022-04-21 12:50:31 signpost: this falls out of the tool being designed as a machine that stamps down a userland wherever you may want one, and not strictly a "distro installer" with strong opinions about what installation means.
asciilifeform: ( along with the removal of the oddball historical wart w/ redundant dirs )
signpost: yeah, so far don't need the /usr symlink. if one's writing a pbuild and it breaks on this path, should fix w/e parameter is demanding the path instead.
signpost: usually this is PREFIX, or something to do with include or lib paths.
signpost: also, looks like right now only things putting anything in sbin are busybox and lilo, so that's easily fixed.
signpost: (b-but those bins are speshul!)
signpost: http://paste.deedbot.org/?id=9nVR << another neat detail imho. the concept of "installed" (which is always relative to a given DESTDIR) includes the specific hash of src+pbuild.
signpost: I've taken a few steps towards binary reproducibility. I'll release this thing before that's complete, but once it exists the build system can be thought of as a key-value store with src+pbuild hash as key, binary output as value.
signpost: at which point there can be agreement among wot for what a canonical binary output of a given pbuild should be.
signpost: were there a DHT atop pest, one can imagine a scenario where one's server fleet requests the binary representation of $HASH, and gets a trustworthy swarm download reply from friendlies.
asciilifeform: signpost: hmm, did we ever find out where gcc ( & gnat iirc ) throw date and random liquishit into binary, making'em irreproducible by default ?
signpost: yeah, found a couple cases. __DATE__ and __TIME__ macros of course, but gnat also craps timestamps of other sorts into w/e those library descriptor files are called iirc.
signpost has a rough patch for latter.
signpost: https://reproducible-builds.org/docs/randomness/ << also this
asciilifeform: signpost: with gnat, trickier, as it has own 'make' system and needs the timestamps on the intake end. they still have no biz ending up in the bin tho
signpost figures that hydra battle can come after release of pentacle. it'll be long.
asciilifeform suspects it's just 1 place in backend that needs patching, there
verisimilitude: This seems to resemble Guix, signpost.
asciilifeform: verisimilitude: iirc guix is a straight binary distro neh
verisimilitude: No.
signpost: I've never used it, nor nix, but would expect many others to have tread this territory.
verisimilitude: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097778 This is like /gnu/store/.
dulapbot: Logged on 2022-04-21 14:04:03 signpost: I've taken a few steps towards binary reproducibility. I'll release this thing before that's complete, but once it exists the build system can be thought of as a key-value store with src+pbuild hash as key, binary output as value.
asciilifeform not tried
signpost: being built atop one's own vpatches means there is no such thing here as "the" distro.
asciilifeform: could refer to the basic package (vtron, booter, etc) as 'the distro' but sure
signpost: but this wouldn't make me hesitate to lift good ideas from others. quite the contrary.
signpost: asciilifeform: yep, there's at least what to bootstrap the thing that need be defined.
signpost building outward from that, solves problems as they're encountered. couple of hundred lines of posix shell atm.
signpost: ought to never be significantly larger than this in the core, nor should core rely on python here, ruby there, brainfuck...
asciilifeform wonders whether the guix people found & patched the gcc random backend turd thing
signpost: deserves a look.
signpost: (btw re: python, this isn't equiv to "python bad, banned!11". if you ask the machine for a blatta, of course python.)
asciilifeform: right, simply banished from center stage
asciilifeform: ( apropos/oblig, asciilifeform's longterm pov re sh & scripting langs )
dulapbot: Logged on 2022-03-17 13:41:46 asciilifeform: mangol: what's imho needed is a fully 'autarkik' scripting lang, which sucks in 0 shitlibs and runs, like 'M', on any extant pc iron under all linuxen from past ~15y
signpost: 100%.
asciilifeform: (through this lens, obv. that pythonism is the reason why no gentoo is ever sitting down in 4MB)
signpost: later, when pestron's had that doesn't demand a python, python oughta slip out just as politely as in.
asciilifeform: the only reason wai e.g. this proggy, or asciilifeform's vtron, or logotron, etc. in py, is that no ecologically clean alternative avail.
signpost: mhm.
signpost: slips out politely if authors are diligent enough to avoid unnecessary interdependency, avoid signing wanton pbuilds.
signpost not sure it's the case, but to contrive an example, say qemu wants python, but can be patched to not.
signpost: (and it's useful without)
signpost: in this case it'd be proper to patch out the dependency.
signpost: but should there be factions that disagree, guidoism can live on under the keys that believe in him. can just distrust those keys, schism happily.
Mocky: http://logs.nosuchlabs.com/log/asciilifeform/2022-03-31#1091821 << I mean, 46 is not *that* old
dulapbot: Logged on 2022-03-31 23:32:14 asciilifeform: send an old man to qatar. comes back broke and with nuffin, let'im starve, then try & find another idiot
signpost: lol!
signpost: howdy Mocky
Mocky: funny that I dispute the 'old' but not the 'idiot', I guess maybe idiot can theoretically improve but old guaranteed to get worse
Mocky: Heya signpost
Mocky: I haven't been keeping up with the latest, just read report of MP's death. Reminded me of TMSR, wondering what other folks have been doing since then, while I've been working on earning my freedom (not yet successfully but making progress)
asciilifeform: oh hey that Mocky ?!
Mocky: same Mocky, yes
asciilifeform genuinely thought fella was dead
asciilifeform: !!key Mocky
Mocky: read the last few weeks of #a logs, quite good discussion 2 days ago re: republic days
asciilifeform: wb to the living, Mocky . and dun feel sad re 'other idiot', asciilifeform was 1st (maybe 2nd) in the queue
asciilifeform: Mocky: didja ever make it back from arabia ? or still there ?
asciilifeform: wilkommen to #a, btw, Mocky . ( and be sure to try out pestnet, to where the action slowly moving )
Mocky: Made it back ok, after 2 months there. Then spent next year getting back on my feet, scraping together funds to pay back my loan
dulapbot: (trilema) 2018-11-15 deedbot: mircea_popescu paid Mocky 1
Mocky: I started to read the pest spec, look a lot more concrete than orig gossipd spec
asciilifeform: was baked to be painfully, pedantically concrete. tho still needs serious work in certain places.
asciilifeform: thimbronion's thing implements ~90% of it
asciilifeform: pestnet contains most of the #a folx, and going for coupla month nao.
asciilifeform aims to bridge dulapnet to pestnet in some sane way, and operate #a as a n00b portal strictly
Mocky: is thimbronion's impl linked from your www?
Mocky: ah, thanks. I'll stand one up, here (Raleigh, NC)
asciilifeform: there's also a c client by jonsykkel ( asciilifeform not yet tested, but iirc shinohai is using it, worx )
asciilifeform: %genkey
asciilifeform: grr wrong term lol
Mocky: lol
Mocky: !!seen lobbes
deedbot: 2021/04/25 01:45:34 <lobbes> http://alt.ericbenevides.com/log/asciilifeform/2021-04-24#1034503 << ty! will do
asciilifeform: !q seen-anywhere lobbes
dulapbot: lobbes last seen in #asciilifeform on 2021-07-31 19:10:38: once re-established a home base of operations, I'll be able to sort some things out
asciilifeform: ^ moar recent
Mocky: gotcha
asciilifeform: be sure to check sig.
asciilifeform must bbl
verisimilitude: Hello, Mocky; this is our first meeting.
Mocky: Ah yes, I think we never have spoken. What are you up to lately?
dulapbot: (trilema) 2019-02-05 mircea_popescu: !!rate verisimilitude -1 children, seen and not heard.
shinohai: $ticker btc usd
busybot: Current BTC price in USD: $40826.82
shinohai: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097847 << using blatta for own station, smalpest for bot atm.
dulapbot: Logged on 2022-04-21 16:58:55 asciilifeform: there's also a c client by jonsykkel ( asciilifeform not yet tested, but iirc shinohai is using it, worx )
verisimilitude: I feel like I got the last laugh there.
verisimilitude: I'm working on my language modelling right now.
verisimilitude: My Latin is still too immature for this, so I'm doing toy versions first, for the most part.
Mocky: I'm guessing that this is a UDP binding for Ada. guessing because it doesn't seem to say that anywhere I could see on first scan, amidst The Raven parody
Mocky: did you craft your comments to all end exactly on the 100 character width?
Mocky: It's not obvious to me what that has to do with language modelling or Latin
verisimilitude: Yes, I do craft my writing so.
verisimilitude: The language modelling is on a different page.
verisimilitude: Also, the UDP actually is relevant to that; I'll explain.
verisimilitude: I'll go ahead and share that I'm working on something which will be like Whitaker's Words, but I'm taking a very different approach to it.
verisimilitude: I realized a ways back my language modelling work gives me a Whitaker's Words style of program with little extra work. It ``falls out'' of the functionality responsible for converting a character stream into the new representation. With that, I just hook it up differently to get the Words program.
verisimilitude: One of the interfaces will be an Ada program for a question-response system over UDP.
signpost: interested to see how this develops.
Mocky: I took 4 years of Latin in high school. Had the (self proclaimed) rare teacher who made us speak it every day
signpost: language itself is a graph, right? not a text-stream, even if we encode the mouth noises that way.
dulapbot: (trilema) 2016-12-21 mircea_popescu: because no, words don't "have meanings". your meanings for ANY WORD are a function of ALL THE OTHER WORDS YOU KNOW. which is why my definitions regularily blow out english dictionaries, wikipedia and other sources of "wisdom" out of the water - i know more words, and in this knowledge i know all the words i know ~better~. infinitely and irreproducibly so.
verisimilitude: I intend to have a toy version targeting toki pona within the month, signpost.
signpost: hell, that was an enjoyable thread, reread.
signpost: verisimilitude: to make mp's statement more applicable to this thread, on the language graph nodes are not strictly whole words. circum-ambul-are for example, as relates to say per-ambul-are.
signpost: will look forward to your item.
verisimilitude: This circum-ambul-are is simply the preposition CIRCVM prefixing the infinitive AMBVLĀRE.
verisimilitude: The per-ambul-are is similar.
verisimilitude: This can be codified.
verisimilitude: However, recall tabular programming, signpost; intelligence or even particular creativity is unnecessary when every valid combination may be explicitly listed.
signpost: I know what the parts of those two words are and mean; do you know what I meant by making reference to them?
signpost: verisimilitude: I find your writing style an annoying barrier, but am I correct that you're saying it's better to have a lookup table for all possible inputs than to write a simple parser for roman numerals?
signpost: (this is bizarre, if so)
verisimilitude: Sure; notice I use a small subset of the full table to get the same effect.
verisimilitude: Consider how many parsers will return 99 for IC, whereas mine never will, unless explicitly told to do so.
signpost: trivially, all programs could also be represented as lookup tables from inputs to outputs.
signpost: this is called caching. what's different in your usage?
verisimilitude: Similarly, my Words program won't parse its input; it will use a trie of every possible word.
verisimilitude: Mine can be trivially proven correct, signpost.
verisimilitude: This also makes it easier to implement such programs in, say, machine code, as there's less to do.
signpost: yes, if you write the correct answers for every possible input in a lookup table, your output will be correct, lol.
verisimilitude: Yes and, through composition rules, those tables can be compressed at the expense of a little extra computation.
verisimilitude: A finite table may then be stretched unto infinity.
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097904 << funnily enuff, this kinda thing was rampant in the 8bit micro days. and not only for e.g. trig functions, but sometimes even (typically 4 bit) multiplication table. faster than 'egypt'...
dulapbot: Logged on 2022-04-21 20:00:38 signpost: yes, if you write the correct answers for every possible input in a lookup table, your output will be correct, lol.
asciilifeform: for extra laffs, sometimes the table entries were ~not~ correct
verisimilitude: Yes; Intel comes to mind.
verisimilitude: http://logs.nosuchlabs.com/log/asciilifeform/2022-04-21#1097894 To walk around and to walk through are the meanings. What was meant by them in particular?
dulapbot: Logged on 2022-04-21 19:42:54 signpost: I know what the parts of those two words are and mean; do you know what I meant by making reference to them?
signpost: asciilifeform: sure, and before this, big books of them.
signpost: caching's a fine thing but I'm curious why verisimilitude is presenting it as a revolution in computing.
signpost: verisimilitude: regarding the latin words, they're conceptually related. would your representation indicate this relatedness?
verisimilitude: This isn't caching.
verisimilitude: Caching is used to confuse performance characteristics. This makes them certain.
verisimilitude: Also, people don't tend to try to apply this to an entire program, only pieces.
verisimilitude: Yes it would, signpost.
verisimilitude: For Latin, I like the idea of allowing implicit concatenation, meaning I like the idea of allowing ``no space'' as one possibility between words.
verisimilitude: For English and other living languages, this would be foolish, but with Latin one can expect a better class of user.
verisimilitude: So CIRCVMAMBVLĀRE and PERAMBVLĀRE could be represented as two words, each.
verisimilitude: I've already thought of this, signpost.
signpost: you're pretty damned bad at explaining things, my dude.
verisimilitude: I could link to another discussion, to help.
verisimilitude: My system clearly needs to be able to express punctuation between words.
verisimilitude: For Latin, one possible punctuation could be ``no space'', to handle exactly these words.
Mocky: verisimilitude: lookup tables and machine code and conlangs... are these in service to solve some bigger problem you'd like to solve or more just "A finite wank may be stretched to infinity"?
dulapbot: Logged on 2022-04-21 20:11:11 verisimilitude: A finite table may then be stretched unto infinity.
verisimilitude: Yes.
verisimilitude: I want to program without text, Mocky.
verisimilitude: However, what passes for text doesn't much suit human languages, either.
signpost: http://verisimilitudes.net/2018-06-06 << here you casually use "better", then indicate that perhaps the category in which you deem "better" is storage efficiency?
verisimilitude: This is the other end of programming without text.
signpost: which if so, sure, you've discovered compression. I think we did this thread.
verisimilitude: Good compression requires a high-level model, signpost.
signpost: hm, so you're atttempting to construct an abstract model of language which you can manipulate with a non-linguistic interface?
verisimilitude: As we've discussed, it's like MIDI, but for human language.
verisimilitude: Rephrase that.
signpost: no.
verisimilitude: Alright.
signpost: lol, asking a cogent question
signpost: *ask
Mocky: Once you can program without text, what benefits will you realize?
verisimilitude: I suppose so, signpost.
verisimilitude: I type less, eliminate categories of mistakes, get a nicer interface, and other niceties.
verisimilitude: These tools will enable me to see more than I otherwise could.
verisimilitude: The MMC makes it very easy to reuse machine code as data, which an assembler obscures.
verisimilitude: An assembler forces the user to remember mnemonics and can only die when facing errors; my tool does neither.
verisimilitude: There are one and two pass assemblers; the MMC is a ``no-pass'' tool.
signpost: the unifying vision behind these is not clear in these statements.
signpost: not saying there isn't one.
verisimilitude: I'll recycle someone else's words.
verisimilitude: ``I think I see the common thread between the two examples you give: a domain mismatch, where the technology used to manipulate a thing is not made of the same primitives as the thing itself.
verisimilitude: A technology that fits its problem domain well can be expected to accomplish two things: first, it should be efficient, by allocating resources only as needed; second, it should disallow elements which are foreign to the domain.
verisimilitude: An example of the latter: a program is conceptually made of things like expressions and statements, not characters; if a tool allows typos to be introduced into a program, the tool doesn't fit the domain well.''
signpost: so I've got another "what it's not", wanton use of bare "should" like we know good things when the lord speaks in our hearts, and a third bare assertion that's also a mess.
signpost: since I'm feeling charitable, one thing you want to eliminate is being able to give the computer invalid inputs, yes?
Mocky: When I look at the volume of your IRC output and the way you drop coupla hundred lines of poetry in front of your UDP binding, and when I imagine how much typing & backspace it must have taken to get dozens of Ada comment lines to all be exactly 100 characters wide, I see that contrary to what you say, your actions suggest that you typing text very highly in the present
Mocky: *that you value, rather
verisimilitude: Yes, signpost.
verisimilitude: I've no choice currently, Mocky.
verisimilitude: I value language, certainly, and this is the way available to me.
verisimilitude: There can be better ways.
signpost: you also wish the computer's state to be as inspectable as possible, thus want programs to (as much as is possible) be tables of enumerated categories of things?
verisimilitude: Yes; code is disgusting where it be unnecessary.
signpost: I don't care about your aesthetic desires.
signpost: but I can see utility in cleanly separating inspectable state from algorithmics
verisimilitude: The less code, the more easily it may be manipulated.
verisimilitude: How many issues arise from scale alone?
signpost: yep. in my example of the two latin words, in my mind the prepositions would be in a table if you like, and the stems another, conjugations yet another.
verisimilitude: When the entire code be a few kibibytes, efficient replacement algorithms may be elided.
verisimilitude: It's not hard to align text so, Mocky; I do that with all of my writings now; view it over Gopher.
verisimilitude: Observe:
verisimilitude: printf '2022-01-31.read\r\n' | netcat verisimilitudes.net 70 | less
verisimilitude: I'm profoundly bored, so perhaps some of this is ``a wank'', but that's irrelevant.
Mocky: I'm trying to see how these pieces fit together for you verisimilitude but it just all sounds so crackpot
signpost: eh the parts I pulled out above make a bit of sense.
verisimilitude: Isn't this channel for crackpots?
signpost: I don't grok how verisimilitude wants to pull them into a useful system, but hey.
dulapbot: Logged on 2022-04-21 22:18:49 signpost: yep. in my example of the two latin words, in my mind the prepositions would be in a table if you like, and the stems another, conjugations yet another.
verisimilitude: It's close, perhaps.
verisimilitude: All words would be in a dictionary; each word would be marked with such information; the declining and conjugating happens at the site of a word's usage.
signpost: only one dictionary, not distinguished by part of speech?
verisimilitude: It depends.
verisimilitude: For AMBVLĀRE, it would be a first grouping verb, and its base AMBVL stored in a separate table.
verisimilitude: For IN, that should likely be stored twice, once as a preposition, and once as the negating prefix.
asciilifeform waits to see what happens when verisimilitude discovers irregular verbs
verisimilitude: Those are stored specially.
verisimilitude: So, ESSE would likely need to be a table storing every mutation explicitly.
verisimilitude: I'm holding off on implementing this until I'm certain I won't run into unforeseen issues, asciilifeform.
verisimilitude: I must learn first and model second.
asciilifeform: verisimilitude: you aint doing this simply for compactness of encoding, i take it. cuz gzip will prolly still, perhaps frustratingly, win (see if can see why)
verisimilitude: No, I ain't.
verisimilitude: I view it as the right thing.
asciilifeform: verisimilitude: what do you aim to do once you have e.g latin encoder ?
verisimilitude: Imagine a learning tool showing ancient Latin, with words coloured and styled by usage, but drawing from this representation; that certainly would beat gzip.
verisimilitude: I intend to play with it, for one.
asciilifeform: an 'objdump' that eats e.g. caesar's gallic wars ?
verisimilitude: I could try my luck bothering unsuspecting Latinists, for another.
asciilifeform: a++
verisimilitude: Yes.
verisimilitude: I just want it to exist.
asciilifeform: fair'nuff
asciilifeform: eggog handling for this type of thing imho interesting problem. (texts on the net tend to be ocr product and fulla typos)
signpost: were you to say more about what it'll do, and less about your feelings, that'd help the model of the item exist in >1 head.
verisimilitude: I try to live my life as if I could die soon; well, I want to die less unfulfilled.
signpost at least has a better picture.
signpost: ah don't worry, you'll always have shit you were going to do the day after you died.
verisimilitude: The implementation and a manual would do better than that, signpost.
signpost: perhaps. I'll read that when it arises.
verisimilitude: I'm glad; I'll try to not disappoint.