log☇︎
42000+ entries in 0.242s
asciilifeform: otherwise mallery would have long ago hired a 'spare' or two, and afaik he never has.
asciilifeform: fromloper: if he gave it out, potentially someone else could bid on the support contract ( under american law, all contracts are theoretically publicly biddable . lockheed et al simply contrive to set up situation where no one else ~can~ perform a given contract )
asciilifeform: the ones who hoard secrets because they imagine that they have a ticket to being part of some future commercial revival of bolix, however -- those are idiots.
asciilifeform: this is a hypothesis. so far my best one.
asciilifeform: re fp, i'll observe that bolix used a weitek fp unit
asciilifeform: this is a q that only he can answer, supposing he feels like answering.
asciilifeform: fromloper: i actually do not know anything re who, if anyone, maintains. aside from phf's admission to having a private fork, linked above.
a111: Logged on 2017-08-21 14:43 phf: i spent (mostly another whisperer and myself did) on getting vlm stable, and i'm unconvinced that some of the issues we encountered were purely "buggy vlm". there is, for example, a crash in floating point instruction that happens when you load document examiner on stock piratebay opengenera. i have no explanation for it still, because vlm code ~seems to do the right thing~. there are other similar instances
asciilifeform: ( a 68k mac.. )
asciilifeform: the 'macivory' did not, used a proggy running on the mac.
asciilifeform: there is, theoretically, a working disasm
fromloper: I wonder if it's going to ask for a license key like Macsyma on the same disk does
asciilifeform: and iirc phf has a working ivory -- worth trying there also
asciilifeform: for all i know, the one and only path to e.g. a room-temperature supercon, was found at one of the reagan scamola firms, and died with it, also. i simply do not know about it.
asciilifeform: ( where i know for a fact that they Had Something )
asciilifeform: to be fair the company was not really a honest commercial co, moar like one of those unofficial usg research institutes, they proliferated under reagan and died with him ☟︎☟︎
fromloper: https://blog.h2o.ai/2008/11/a-brief-conversation-with-david-moon/
asciilifeform: weinreb died a few yrs ago
asciilifeform: funnily enuff a Dave Moon ran for election to some city dogcatcher post , not far from where i live. but on examination seems to be different d00d.
fromloper: this document refers to Dave Moon's files several times; I've tried to find any public presence of Moon on the net, maybe a mail address - found nothing
asciilifeform: somebody , maybe same old man, has a bookcase full of typewritten design apocrypha, memoes, etc. from ivory group. he ain't sharing either.
fromloper: yeah, somehow a lot of potentially interesting people on social media end up mostly posting this kind of crap ☟︎
trinque: pretty common to end up LARPing in retirement, even if one lived as a man, unfortunately
fromloper: there is a whole section on the virtual memory
apeloyee: a hash of the entire tree doesn't take much
asciilifeform: fromloper: i'll confirm, this is a larger and moar detailed document than old one.
apeloyee: I proposed being able to name arbitrary required antecedents << also probably needs a mechanism to declare "there are no other files in the tree" ☟︎
asciilifeform: i do not have, nor ever had, a working 'ivory' of any description.
apeloyee: "may be a 50kg sword" << doesn't seem to be. can be retrofitted into an existing design. as i said above "there needs to be a tree hash in the _leaf_ patch. and it MUST match the resulting tree"
asciilifeform: apeloyee: trinqueian / mircea_popescuine vtron is arguably The Right Thing. my observation is that it may be a 50kg sword.
apeloyee: when you sign a tarball, the signature is not transferrable to anything else
asciilifeform: apeloyee: there is not a mechanical solution to preventing someone from 'putting in format c:' proverbially
asciilifeform: fromloper: it is not a complete arch description, you cannot write a working emulator with it ( or even make the existing snap4 not-crash )
trinque: then we are closer than it appeared in the long thread. I proposed being able to name arbitrary required antecedents in a vpatch's header, and this appears equivalent in effect to copying the file in whole.
trinque: asciilifeform: to see if I can restate your opinion back to you, if I edit (as single author) both readme.txt and doesallthework.adb in separate vpatches, your view is I combine those into a single vpatch, if I want to build atop both in a new vpatch?
asciilifeform: i don't have a good counter-argument to this.
asciilifeform: and in both cases, the ability to explicitly mark subsystems as independent ( e.g. a readme.txt being independent from doesallthework.adb ) is lost.
apeloyee: well, a hash is not the same as the signature, but otherwise yes.
a111: Logged on 2018-01-17 19:31 apeloyee: a vpatch's purpose is twofold. 1) to provide a way to construct some files based on some antedecent files, whose hashes are given. 2) to take some responsibility about the entire tree. but the signature on a vpatch doesn't fix the state of the tree; it is defined implicitly by antedecent patches, which are liable to change at any time ("regrinding") and thereby change some files not...
apeloyee: that's a spurious objection. one need not to sign an antedecent state, one needs to sign a RESULTING state. to expand on http://btcbase.org/log/2018-01-17#1771900 , you're free to pick individual files from wherever, possibly several different trees, but there needs to be a tree hash in the _leaf_ patch. and it MUST match the resulting tree (under the principle that patch author takes the... ☝︎
BingoBoingo: The closest thing so far was they time an Israeli embassy worker was busted with a bomb or fake bomb near the WTC
BingoBoingo: And in other news, a police car exploded near the Brazilian border https://www.elobservador.com.uy/bomberos-explosion-auto-policial-fue-intencional-n1160744
apeloyee: well, your point seems to be specifically that work which can be done by machine is shifted onto a human. this is insane.
asciilifeform: apeloyee: trinque and mircea_popescu would like to put more of it on the machine. i haven't with what to dissuade them, it is a philosophical q, not even technical.
apeloyee: files are NOT INDEPENDENT. despite CVS and v pretending they are. this is a problem. you could have required some form of cryptographic commitment to either the tree state or even the antedecent patches themselves, but didn't ☟︎
apeloyee: ...referenced in a particular vpatch
apeloyee: a vpatch's purpose is twofold. 1) to provide a way to construct some files based on some antedecent files, whose hashes are given. 2) to take some responsibility about the entire tree. but the signature on a vpatch doesn't fix the state of the tree; it is defined implicitly by antedecent patches, which are liable to change at any time ("regrinding") and thereby change some files not... ☟︎
BingoBoingo: Now that was a wasted voice
Covale`: http://btcbase.org/log/2018-01-17#1771610 - no, not mine, I don´t have one. Neither do I have that cup, but it´s a pretty decent Lenin nonetheless ☝︎
asciilifeform: i can't picture how without a megatonne of conditionaljumps. but maybe just failure of my imagination.
apeloyee: from above: "Check that the actual parameters of a subprogram call are not aliases of one another. To qualify as aliasing, the actuals must denote objects of a composite type, their memory locations must be identical or overlapping, and at least one of the corresponding formal parameters must be of mode OUT or IN OUT. "
freetlas: asciilifeform: Just a person who likes to read trilema from time to time :)
apeloyee: https://gcc.gnu.org/onlinedocs/gnat_ugn/Alphabetical-List-of-All-Switches.html#Alphabetical-List-of-All-Switches : "-gnateA" Check that the actual parameters of a subprogram call are not aliases of one another.
asciilifeform: speaking of which, could even have a stackmachine with circular ring for a stack. ☟︎
apeloyee: it's a wonder -gnatp doesn't disable nullity checks
asciilifeform: the first cell has a false HasPrev ; the last -- a false HasNext.
asciilifeform: so a stack cell would contain not only an FZ of the current bitness, but two boolean values, e.g. HasPrev and HasNext
asciilifeform: you can trivially show that any attempt to walk under or over the stack, would have to involve a null-dereference.
asciilifeform: want() would then vanish; both stack underflow and overflow checks would be handled by the nullity check ( first cell has a null in its 'prev' slot; last cell in stack -- in its 'next' . )
asciilifeform: apeloyee: here's another idea from my notes , that would do the job : to dispense with the array representation for the stack, in favour of linked list. ada permits the definition of a 'not null' pointer type (whose non-nullity is checked on every reference) .
asciilifeform: one way to model this process is that there is an 'ugliness budget', just like there is a cpu cycle budget, that can be 'spent' in certain ways
asciilifeform: how about when it goes in a 25mhz micro ?
asciilifeform: this is a fundamental headache
asciilifeform: pc arch also does not, unfortunately, give a way for userland to trap on reads/writes to specific piece of memory
asciilifeform: and at any rate the right place for such a thing is in the code, rather than in the guts of the linker
asciilifeform: there is not, afaik, a way to force stack frames to be explicitly padded
apeloyee: put a dummy array of 4 (or whatever) elements just before the Stack
apeloyee: doesn't gnat have a facility to control the layout of memory?
asciilifeform: mine is also ugly, it is conceivable that somebody, some day, in a broken variant and running with -gnatp, ~will~ write to the zero cell
asciilifeform: the fact that an array is indexed by a type which has a range outside of the array's.
a111: Logged on 2018-01-17 17:03 apeloyee: the proper range for a cursor into an array (1..N) is (0.5 .. N+0.5); this is usually shifted to become (1 .. N+1) as in text editors, but no reason not to shift in the other direction, as you did, to (0..N)
asciilifeform: this is not equiv to cursor behaviour in text editors, though. there, cursor always is pointing to a valid fillable cell
apeloyee: well it'd turn a pop from empty stack into a range-dipping eggog instantly << the reason I even suggested that 3 weeks ago
apeloyee: "1) ugly" << can't see that.; "2) ... it relies on type ranges for good chunk of the proofolade"<< if you really want, can explicitly declare a subtype of Stack_Positions, omitting 0 from it
apeloyee: SP _is_ a cursor, I'm merely suggesting to treat it as such
apeloyee: the proper range for a cursor into an array (1..N) is (0.5 .. N+0.5); this is usually shifted to become (1 .. N+1) as in text editors, but no reason not to shift in the other direction, as you did, to (0..N) ☟︎
asciilifeform: well it'd turn a pop from empty stack into a range-dipping eggog instantly
apeloyee: an extra element won't save the father of russian democracy (c), if FFACalc stack manipulation code is wrong; e.g. if a 'Want(X)' statement is omiitted
apeloyee: Logically, SP isn't a reference. It's a cursor, showing the boundary between the valid and invalid elements. If a line in a text editor has N sybols, then a cursor has N+1 valid positions. (Consider that an empty line sill has 1 valid cursor position).
asciilifeform: apeloyee: a long-term goal is to have the whole proggy bulletproof even after building under '-gnatp' ('disable all range checks'). relying on the range check of Stack , vs SP's range, does not go well with this
asciilifeform: i.e. in all cases where the array has a custom indexing type as the index, the array exists over the entire range of said type.
BingoBoingo: In other shithole factories: meltdown/specte patches are making a bunch of industrial systems wobble! Chinesium's about to get a low more Pinoy
apeloyee: can you think of a way to have the range of SP and of Stack be the same << this is plainly absurd. a N-sized stack has (modulo contents) N+1 possible states: "0 elements", "1 element", ..., "N elements". i.e.the ranges MUST differ, by exactly one.
asciilifeform: can you think of a way to have the range of SP and of Stack be the same, but to still represent concept of 'empty' ?
asciilifeform: proggy ought to be written in such a way that the reader can ~see~ that it is correct.
asciilifeform: thing is, a sparkism is not a substitute for a 'fits-in-head'-correct routine.
asciilifeform: possibly one can make a SPARKistic proof of non-aliasing , for proggy taken as a whole. i'ma look into it.
apeloyee: well, ideally it would have a "No_Out_Arguments_Aliasing" restriction which would insert runtime checks ☟︎
asciilifeform: and afaik gnat does not know how to prevent aliasing in the general case, but only in a few specific situations ( array copies )
asciilifeform: it isn't a wrecking
asciilifeform: ffa itself is a sort of tightrope walk, an attempt to 'и рыбку съесть и нахуй сесть' . it isn't actually possible to make all of the routines able to take every conceivable kind of compile-time abuse ( which i'd argue overlap of arguments, is ) without conditional jumps.
apeloyee: does gnat provide a facility to check aliasing at runtime? ☟︎
asciilifeform: the 1 caveat re buffers, is that there is a practical limit as to what can be made anticlobbering: if i were to do it to items that occur in inner loops of O(n^3)istic items, e.g. mux, proggy will end up 'geological' 4ever.
apeloyee: and add a new exercise: re-read all previous chapters and write what aliasing of arguments is safe
asciilifeform: this is a bit of a downer, i confess that i hoped apeloyee had found a more subtle, lethal boojum.
asciilifeform: because there isn't actually a limit as to what less-clueful people can break, regardless of what i do
esthlos: I will proceed one step at a time
asciilifeform: has apeloyee found a specific instance where it can be made to eggog ? or is this a hypothetical 'once less clueful people start changing things' observation ? ☟︎
asciilifeform: though it is a waste of space and time.
asciilifeform: prolly FZ_Mod_Exp oughta accumulate Product in a temp, and shit it out in 1shot in the end, like FZ_Mod_Mul.