313100+ entries in 0.191s
trinque: phf: eh nevermind, found a
typo :p
pete_dushenski: inb4 stan makes his
token comment about
the 'fundamental retardation of accounts, which must die (tm) (r)'
trinque does it one more
time, for science.
trinque: I must have filthy python habits,
thought it'd know when
the lisp was written later
trinque: phf: how do I keep my fasl from being out of sync with changes
to
the lisp ?
trinque: pete_dushenski:
try
teh $up again
pete_dushenski: mircea_popescu: we'll see if
this 15 minutes of fame lasts longer
than my stint as a lord ;)
trinque: I'm fixing some issue with gpg
that's messing up $up
mircea_popescu: check yourself out pete_dushenski you get
the #trilema log global first!
phf: yeah,
there are some very obvious stats
that can be ran, including wot analysis, who speaks
to whom, frequencies, etc.
mircea_popescu: logging
teh forum happens
to be a field where a lot of interesting
things can be done. right off
the
top of head : you familiar with pisg ? ; something like an interaction graph based on namecalling may be interesting.
phf: haslink is actually only ever mentioned 8
times in log
mircea_popescu: providing search for corporate clients. if you get yours working as well as his, perhaps can be arsed
to market it.
mircea_popescu: fwiw, mthreat,
the guy who was providing
the search, actually did
that as part of a start-up
that got sold off or spun off or w/e
they do.
phf: no, but i can add
that, search is entirely reworked, since it's essentially grep by
text only, so i'll have
to refine it bit by bit anyway
phf: which is just immediately next
t-a message by
timestamp
phf: i'm going
to deploy
to
the same machine as btcbase, so
the log url is going
to be btcbase.org/log/
phf: yes,
though not
tonight
mircea_popescu: phf i dun recall what you answered, will i be able
to update log references in
trilema with a single sql statement ?
phf: right now
the switch over between
t-a and b-a happens at log entry 1440677, which is "mircea_popescu: and so,
this is
the end for
this particular irc channel as a venue for
tmsr. ..."
phf: ok, logs are at a reasonable point, going
to go for food and walk and
then do a deploy
mircea_popescu: tho honestly a) i see no difference between "modernity" and "lazy stupidity" ; b) i see no clear reason high clock speeds are necessarily related
to stupidity.
mircea_popescu: hey, i have no argument with
the "dram is a pain in
the ass
to do hands on stuff
to"
mircea_popescu: right. exactly in
that sense. you'll have
to feed it anyway./
mod6: what does alf's dream
turing machine look like?
mircea_popescu: through a similar decision process as you propose, rats are not superior
to cockroaches
mircea_popescu: you can either have
this chip be 256 mb or 4gb. what do you prefer.
mod6: there has got
to be an individual cost, but it's probably rolled up into single "unit" of storage.
mircea_popescu: "higher cost per bit" is significant
tho. it's either
two capacitors, or
two capacitors +
two diodes
mod6: Yeah, kinda looks like it. (Was just
taking a peek
there...)
mod6: back when
that was a
thing it probably was more affordable for Mr. Luser. What about
these days?
mod6: asciilifeform had a interesting dream
too haha.
midnightmagic: it *can* be meaningful in
terms of signal propagation and logic locality.
mod6: asciilifeform: intresting
though.
mod6: oh oh, we're ourselves on
two seperate branches of conversation.
mod6: they're lying
to me
mod6: The other
technique is
to write programs without branches, or with fewer branches,
typically using bitwise operations instead. [1] [Knuth, Donald (2008).
The Art of Computer Programming. Volume 4, Pre-fascicle 1A (Revision 6 ed.). pp. 48-49.]
mod6: oh. right, can't you avoid
them with bitwise ops?
mircea_popescu: what he was
talking about was
that in all fpga/asic designs,
the branches are
the
time waster / heat dissipater
mod6: if not, maybe a word or 6 on what you're getting at
there.
mod6: oh do i understand you correctly on
the branching problem -- if
the stackframes are signed, how will i know (in advance) which address
to jump
to if all addresses are assigned at run
time?
mod6: heheh. i'll have
to scroll back.
mircea_popescu: pwnm (tm)
thought your dream is eerily accurate and
true
to life.
mod6: was weird. maybe i
took
the acid.
mircea_popescu: that sort of
thing only "works" if patently nonsensical, as per
the original.
mod6: you know what i'm
trying
to say?
mod6: i dunno, on some weird paranoia
tangent of
trying
to "only execute what alf said was good."
mod6: a register before executing
the next op.
mod6: is
there anything
to my wack-o dream? could a method be developed
to have signed stack frames
that
the cpu would only execute up successful validation? would
this even be worthwhile? probably not, a guy can still probably grab
the IP and point it where he likes. probably would have
to have some sort of pipeline (is
this a
thing?)
that would make sure
that
there was valid checksum (from signature) held in
mod6: he was going
to do
this V like
thing where each op has a corresponding hash, and at
the end of all
the ops,
the hashes had
to match or
the cpu wouldn't execute
the
thing...
mod6: i had
this dream lastnight
that alf called me and wanted
to
tell me about
this new method of processing... but i couldn't hear him very well as I was going
through
this
tunnel.
mod6: the pics of
the rigs
these guys had of
their GPU farms kill me