asciilifeform: imho v is the simplest, mechanically, versionatron ever baked. so i dun expect Mocky will have much of problem
asciilifeform: ( even tho it is about the ancient proof of concept vtron, with old-style hashes etc., re the basic mechanics it is still 100% correct )
asciilifeform: ( incidentally, 'can redefine cpu instructions in boot rom for custom kompyooting' dun require fpga etc. fancy modern tech, e.g. dec alpha had it )
asciilifeform: likewise, operator at least oughta have the ~option~ of 'all MULs take same # of cycles', even if no one wants this on server.
asciilifeform: the 'meta-sanity' arguably is that operator oughta be able to at least switch off the cache. ( you cannot, e.g., on x86, switch off l0 cache )
asciilifeform: ( to take simple example, you defo want cache on server, but it is absolutely lethal on ciphertron )
asciilifeform: theoretically what is 'sane' for machine to fiddle with proggies on, is not appropriate for server, or for ciphertron, etc
asciilifeform currently suspects that '1 sane comp for all problem domains' is a misapprehension of the problem, and that '1 for crypto, 1 for other things' is merely the beginning of it.
asciilifeform: if nuffin else, some of asciilifeform's demands re 'sane comp' are in architectural tension with others ( rather like to ask 'i want sword that cuts other swords, but isn't brittle' )
asciilifeform: http://btcbase.org/log/2019-03-26#1904962 << not to let this escape; i'd actually be quite interested to read diana_coman's own thoughts re what is a sane comp. ( asciilifeform wrote at great length re the subj, would also like to see where folx disagree / expand )☝︎
asciilifeform: but prolly cannot escape to have a glue for it, while we sit on irc , load www for various uses, etc
asciilifeform: rather than reasonable foundation for anyffin new
asciilifeform: asciilifeform sees tcp as a legacy tech, really
asciilifeform: and not only wwwistic, but e.g. irc bot
asciilifeform: i'll add that even a tcp skin wouldn't be entirely useless ( right nao the only way to write a wwwistic proggy in ada is to use adacorpse's 'gnatsockets' crock of shit )☟︎
asciilifeform: then can talk to mysql etc. just like trinque's cl proggy does
asciilifeform: mircea_popescu: not at 'pen' stage , but prototyped, iirc i posted an example with trad btc tx.
asciilifeform: ( in the last iteration , the 1 that presently dun build on acct of gnat bug , you simply make any data structure you like in ada, and it persists to disk. )
asciilifeform: asciilifeform's mmap thing (ideally a working ver of the latest iteration, but even the original 'horsecocks') already gives ability to write simple , fast db, for uncomplicated schemas
asciilifeform: ideally what you'd want imho is a sane db solidly in ada, rather than coupla mil+ loc of c ??? . but this may be bridge too far just yet.
asciilifeform: mircea_popescu: as i understand , this one would need either tcpism (not written yet) or unixsocketism (also afaik not written, aside from the sad adacorpse implementation in gnat std lib)
asciilifeform: the 'bootstrap to c + linux' people make this exact mistake. 'look, i wrote this compact bootstrapper, nao i can breathe out and start wallowing in my familiar pointerolade an' nulltermstrings pigsty again'
asciilifeform: there was an old mircea_popescu piece , where , 'some people - stupid in particular way, and think that if they buy big enuff house, it will never have to be cleaned'
asciilifeform: imo peh is considerably closer to a tabletop model for sane iron cpu, than to bootstrapper for x86 etc
asciilifeform: i would not presently go as far as to say 'peh is ~the~ bootstrapper' ( current peh cannot even output a raw byte... ) but suspect that the solution will have approximately its shape.
asciilifeform: no good reason imho to try to keep c on life support in long term. it's done enuff damage to the field.
asciilifeform: ( peh per se is a 'peh compiler', arguably, in that it can and does produce peh tapes as output, this is how e.g. keygen worx )
asciilifeform: mircea_popescu: it is entirely conceivable that someone may want to bake 'peh compiler' . but intent of design is specifically for it to be entirely usable without any such thing, with reasonable effort.
asciilifeform: ( in fact ada per se has a fascism knob that, if set, prevents early return from subroutines . i have not thus far used, cuz in ffa per se this is already the case, nuffin gets to terminate early when 'constant time' algo )
asciilifeform: imho it's not only worth doing, but absolutely must be done if yer working with reasonably compact pieces of safety-critical proggy
asciilifeform: note that just about any iterative algo can be rewritten in this form. ( just that ~nobody ever bothers , typically )
asciilifeform: all of this, is to 1) make easier to 'fit in head' the flow of a pehtape by eyeballing 2) reduce cachemiss sidechannel chatter
asciilifeform: all subs have 1 valid termination point, there is no early 'return'; ditto loops ( as of ch18 )
asciilifeform: in re peh in particular, the 'tape' model i picked consciously, and not simply in re the terminology to call the proggies 'tapes'. thing is built so that all motion is , albeit non-monotonically, forward ( e.g. subroutine can call another sub, but only one defined ~earlier on the tape~ from the place where called; and all subs ~must~ either reach their termination point, or produce verdict & halt the process , there is no 'goto' )
asciilifeform: 'random access' is a very expensive illusion .
asciilifeform: mircea_popescu: right. and , interestingly, the coupla-1000fold diff b/w main ram speed and that of cache , on current-day irons, has just about brought back the age of tape algo
asciilifeform: ( recall incidentally the fate of 'bubble memory.' mega-invention, btw. d00d discovered that you dun need a tape deck to use tape, in a certain config of magnetic field, you can make the bits on tape ~walk~ in circles. and r/w'em as they move past a particular spot. )
asciilifeform: all of this being said, i suspect mircea_popescu would still barf if to e.g. fetch a trilema pg, 5km of tape had to move a full circle.
asciilifeform: you can do a surprising amt with N tape decks that ordinarily 'demands randomaccess'
asciilifeform: where he had the 'algos for N tape decks'
asciilifeform: mircea_popescu: re 'do you really need arbitrary access so often' : recall knuth's 'centerfold' page in vol.2 of aop
asciilifeform: or a 2pass compiler. or anyffin that gotta walk a substantial length of bits >1ce
asciilifeform: mircea_popescu: per the current instruction set, you could not write e.g. 'unzip' in it
asciilifeform: thing is really meant for sequential work on relatively compact items
asciilifeform: mircea_popescu: there's other constraints, e.g. peh is quite certainly not optimized for random-access on large data set , there is no support for even such thing as array
asciilifeform: bvt: not merely 'sad result'. sad ~objective~. recall, it was his ~objective~ to bring up gnustack.
asciilifeform: problem is that you eat at least 256bit to add 1+1
asciilifeform: oh thing per se could easily fit in coupla kb
asciilifeform: somewhat heavy ram-wise for a general-purpose script lang tho
asciilifeform: funnily enuff, e.g. ch18 peh actually worx as 'a php' (i.e. can do string substitutions and elementary arithmetics... )
asciilifeform: mircea_popescu: the 'bootstrap' concept in that thread was specifically re 'what coupla kB can you start with in rom that would actually let you build the entire os', rather than the traditional 'boot' process , not to be confused with
asciilifeform: bvt: the other thing, is the 3-ring circus aspect of elaborately dethompsonizing a box in order to... bring up 1M+line of linusolade☟︎
asciilifeform: bvt: imo c is intrinsically broken ~conceptually~ as a bootstrapping lang ( entirely aside from how it is broken as a ~platform~ lang, re which subj asciilifeform has entire www... ) -- in that it intrinsically demands a massive pile of nickel and dime utils ( bash, 'make', etc ) to work as usable platform
asciilifeform: at least Framedragger was under the notion that he'd get to fuck mermaids at the bottom of his sea, or what was it.
asciilifeform: pretty sad, imho : esthlos wrote a++ log summaries, a working vtron, possibly other items i cant recall
asciilifeform: bvt: seems like mircea_popescu's original eyeball verdict was 100% on target? i.e. 'mes' is a pile o'shit masquerading as they usually do for a solution
asciilifeform: re 'in congress directly' -- anyone recall the congressderp who had zombie-esque blue skin from taking 'colloidal silver' patent medicines ?
asciilifeform: incidentally '50s usa had own version of term for 'aspirational product'. popular decoding of auto brand 'pontiac', famously, was 'poor old nigger thinks it's a cadillac'
asciilifeform: mircea_popescu: they moar or less ~are~ 'in congress directly'
asciilifeform: 'aspirational product'(tm) or how did tlp put it
asciilifeform: i can't picture how to argue that c ~wasn't~ a socialist
asciilifeform: mircea_popescu: i pictured '1947 'red' not as kennan et al tho, but as e.g. patton, churchill, et al - the folx who were ready to rearm germany an' proceed straight into ww3, and had to be stfu'd by the brass strictly on acct of truman having insufficient nuke pile (iirc atm ~7 or so)
