#asciilifeform | 2021-08-18

lol

~ 10 hours 51 minutes ~

14:20

cgra

http://logs.nosuchlabs.com/log/asciilifeform/2021-08-17#1052829 << live approximately within a few days, will announce here

↖

14:20

Logged on 2021-08-17 14:39:42 billymg: cgra: glad to hear, is your blog live anywhere yet? and definitely let me know if you run into issues or see things you'd like added

14:26

cgra

http://logs.nosuchlabs.com/log/asciilifeform/2021-08-17#1052830 << i currently have other plans, but not 100%, so i will keep this in mind, thanks!

14:26

Logged on 2021-08-17 14:51:37 asciilifeform: oblig. spam: would be honoured to offer cgra rack service! supposing he doesn't already have.

~ 1 hours 10 minutes ~

15:36

*

asciilifeform looks fwd to reading cgra's new www

~ 2 hours 14 minutes ~

17:50

!w poll

17:50

Polling 17 nodes...

17:50

84.16.46.130:8333 : Could not connect!

17:50

185.163.46.29:8333 : Could not connect!

17:50

205.134.172.26:8333 : Alive: (0.082s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=696405

17:50

54.39.156.171:8333 : (ns562940.ip-54-39-156.net) Alive: (0.110s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=696405

17:50

205.134.172.6:8333 : (172-6.core.ai.net) Alive: (0.089s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=696405

17:50

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=696405

17:50

71.191.220.241:8333 : (pool-71-191-220-241.washdc.fios.verizon.net) Alive: (0.160s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=696405 (Operator: asciilifeform)

17:50

205.134.172.27:8333 : Alive: (0.145s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=696405 (Operator: asciilifeform)

17:50

205.134.172.28:8333 : Alive: (0.144s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Return Addr=0.0.0.0:8333 Blocks=696405 (Operator: whaack)

17:50

208.94.240.42:8333 : Alive: (0.159s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=696405

17:51

143.202.160.10:8333 : Alive: (0.234s) V=70001 (/therealbitcoin.org:0.7.0.1/) Jumpers=0x1 (TRB-Compat.) Blocks=696405

17:51

54.38.94.63:8333 : (ns3140226.ip-54-38-94.eu) Alive: (0.261s) V=88888 (/therealbitcoin.org:0.8.88.88/) Jumpers=0x1 (TRB-Compat.) Blocks=696405

17:51

213.109.238.156:8333 : Alive: (0.336s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=696405

17:51

185.85.38.54:8333 : (tlapnet-38-54.cust.tlapnet.cz) Alive: (0.294s) V=99999 (/therealbitcoin.org:0.9.99.99/) Jumpers=0x1 (TRB-Compat.) Blocks=689331

17:51

176.9.59.199:8333 : Violated BTC Protocol: Bad header length! (Operator: jurov)

17:52

192.151.158.26:8333 : Busy? (No answer in 100 sec.)

17:52

103.36.92.112:8333 : Busy? (No answer in 100 sec.)

17:52

$ticker btc usd

17:52

busybot

Current BTC price in USD: $45895.23

~ 2 hours 57 minutes ~

20:50

whaack: your subscription renewal is officially processed! congrats! you're paid up nao through 16 aug 2022 !

20:50

Logged on 2021-08-17 14:36:53 whaack: i tried to verify the encrypted msg

21:03

whaack

asciilifeform: awesome!

21:11

*

adlai spent nearly an entire circadian tweaking secondhand irons, to get a better idea of the labors that would be saved by an RK server subscription

21:11

Logged on 2021-01-19 19:14:37 adlai: is seriously considering, in case this was not crystal clear by now, 5th and 6th decades of life on this green earth as electronics junkyard proprietor.

21:13

I realise that the service S.ALF is providing is has almost entirely null intersection with the services of 'junkyard proprietor'; however, in this specific case, my work is towards a 'home server' setup

21:15

*

asciilifeform realizes that adlai is likely speaking figuratively in re 'S.xyz' , but must point out for readers that asciilifeform's isp does not solicit or accept investments, a la mp's operations or in any other way other than via custom.

21:16

http://logs.nosuchlabs.com/log/asciilifeform/2021-08-11#1052263 << I have a few specific questions, after reviewing this... the most pointed one is whether the "cleanup charge" is the only moving part in the accounting that depends on bandwidth consumption.

21:16

Logged on 2021-08-11 22:08:03 asciilifeform: adlai: so then, read the instructions and see which if any catalogue item worx for you. (plz note that the prices are not negotiable)

21:16

( see also possibly-entertaining vintage wank re subj )

21:16

Logged on 2019-11-08 20:47:52 asciilifeform: if i operate a parking garage, and i permit obama to park his abrams tank, it does not constitute a '20mil $ investment in asciilifeform's garage' .

21:16

adlai: plox to link to selection from the handbook article

21:17

HFT could be considered spam, and in fact some exchanges use the words interchangeably.

21:18

broken link

21:18

blyart

21:18

it's to the paragraph containing "release of an IP address ... will incur a 50 $ cleanup charge."

21:19

adlai: this simply means that if you've rendered an ip addr unusable, but intend to continue subscription, you will need a new one, and this is a considerable bother on my end that is to be paid for.

21:20

fwiw, the rate limiters that I've seen in the exchanges that I've actually used vigorously enough to have experience with, rather than merely having read the terms of their API, all reset after e.g. 24 hours

21:20

ftr this has not, of yet, happened to any of asciilifeform's subscribers.

21:20

so I do not anticipate wrecking address registrations

21:21

adlai: the intention of that paragraph is to remind subscribers that asciilifeform aint in the biz of supplying 'fresh' ip addrs to replace 'burned' ones

21:21

the addrs themselves are leased from upstream vendor and relatively expensive.

21:23

right. my pointy question, however, stems from wondering whether the contract and billing described in the link are fair for a server planned to be used for a tradebot that runs as close to the ratelimits as practical; and my concern is that my server would be an outlier in bandwidth consumption.

21:23

adlai: what kinda bw consumption do you anticipate ?

21:24

(avg. case)

21:24

I have no sane and honest ballpark.

21:24

adlai: may want to estimate. entire cage runs off a 100mbit pipe. (at the moment aint particularly crowded, but this is the peak bw avail.)

21:25

the cliche about "bandwidth consumption by streaming porno and zoom circlejerks grows to fill the available capacity" probably applies to automated trading systems, too.

21:26

adlai: pretty sure i explained this previously, but the current rule is, if yer 'a hog' and other subscribers complain, you will get to pick b/w being dead last in QoS prioritizer or to be billed for segregated personal pipe (~100 usd / mo)

21:26

it's probably saner to be aware of the cage's bandwidth capacity, and chop downwards from there to compute soft sanity and hard insanity limits of bandwidth consumption, rather than to bloviate about a configuration that I've never tried yet for my program.

21:27

those options seem fair, especially considering that nobody has yet caused such complaints.

21:32

fwiw, my current 'rate limiting' is mostly locking inefficiencies and various calls to (sleep (/ small-integer)) ; when these are all removed, program consumes the public firehose from the exchange, and sends position adjustments as rapidly as the exchange allows

21:32

obviously, not a tactic employable by any significant fraction of the exchange participants, otherwise there is a ridiculous blowup in communications complexity

21:33

adlai: outta curiosity, made any dough yet doing this ?

21:33

*

adlai suspects that some theoretical result along the lines of "bucketing improves market efficiency" is provable, although it is not exactly politically palpable research to publish!

21:34

ya know how it goes, no?

21:34

no?

21:35

the kind of trading I do is epsilon-profitable almost all the time, and gets sporadic coarse losses every now and then.

21:35

soo, profitable if you don't count the losses? lolk

21:36

the kind of market crashes that happened around mid-february and -march of 2020, and the kind of crazy rallies that happen in bull runs, are too rapid for a naive volatility harvester, and cause losses.

21:36

from isp pov aint any of my biz. but as someone who silently dabbled in autotraderism in past years, cannot resist to ask.

21:36

no strategy is ever 'profitable'

21:36

all are probabilistic, until you are examining a sequence of trades already performed in a market, and then it either made money, or lost money.

21:37

*

adlai has accounts that were net-profitable, over the course of an entire trading experiment... if you account only in fiat value of assets!

21:37

adlai: ~that~ is as easy as 'hodl', lol

21:38

iirc, the only account that was ever net-profitable in btc over the entire experiment was S.MPEX

21:38

*

asciilifeform however considers 'profit' to be when you end up with moar of the unit you had put in, at the end, than at start.

21:39

right

21:41

amusingly enough, the code that I had written up to the time of starting the S.MPOE experiment had this kind of logic implicitly hardcoded into it, and as a result, spent a disturbing amount of time accumulating its profit in shares

21:42

it was not difficult to change the logic so that this bias was configurable, and then configure it to keep profits in btc instead.

21:43

adlai: i suppose was a++, while pyramid was going

21:43

Logged on 2021-08-10 20:34:17 asciilifeform: even a rube like asciilifeform , even in '12, was able to see that it was a pyramid

21:43

*

adlai wonders whether to comment here, or on the article, that payment issues need clarification now that deedbot's wallet service is gone

21:44

adlai: here plz.

21:44

adlai: see here 1st.

21:47

*

adlai hopes that this conversation does not count as negotiation, nor "concerning the details of a payment"

21:49

adlai: at the risk of pedantism -- if you pay to an addr that you saw somewhere other than a signed message from asciilifeform specifically concerning your subscription -- you paid to /dev/null.

21:49

on top of this (and this is in handbook) you 1st request the machine, then asciilifeform confirms that it is in stock, installs os, plugs it in, ~then~ you pay.

21:50

then you get a signed msg w/ login & ssh fp.

21:50

e.g.: is it sane/polite/reasonable to lease either the 256G or 1T options for a trial month, so that the locked-in price of a subsequent long-term lease can be for the 128G in case that is sufficient.

21:50

*

asciilifeform thought this was pretty clear from the written instructions

21:51

my gut says that this is an unreasonable approach, because I know that even 128G is multiple OOMs beyond what I require, and the extra gigs are arguably ye olde 'problem of too much money'

21:52

"oh, I have all these extra gigabytes and don't have to throw away live trade feeds and orderbook data every day? I know, I'll train neural networks!"

21:52

adlai: have you used rk before ?

21:52

asciilifeform: your pedantry is, as usual, reasonable.

21:53

nope.

21:53

adlai: may want to make sure your softs work on arm64.

21:54

(e.g. sbcl does. but if you have oddball custom extensions, then nfi)

21:54

adlai: anyways lemme know what size of ssd you want, and pgpgram to asciilifeform ; i may have to order it.

↖

21:55

*

adlai hypothesizes that, on the continuum between oxen and chickens, the RK's horsepower is... four velociraptors?

21:55

(trilema) 2018-11-11 asciilifeform: ( asciilifeform strongly suspects that the idea of replacing '2 strong oxen with 1024 chickens' in the ~general case~ is a perpetuum mobile. witness the heathens and their 'cloudism' , it is a circus )

21:56

adlai: e.g. jurov's lxr, dpb's blog, coupla other people, currently on rk

21:56

asciilifeform's logger was on a rk for most of 1st yr post-piz

21:57

iirc billymg's logger is on a rk

21:57

I doubt that any of the parts that are not standard CL are finicky enough to care whether it's arm64 or x86_64

21:57

*

adlai has encountered headaches with 32bit machines, though!

21:57

it's an arm64 adlai

21:58

(fat lotta good it does tho, w/ only 2G of mem)

21:58

(there was a 4GB variant but they were historically ~unobtainable)

22:00

the 2G RAM is an interesting challenge; I currently float between 100M and 1G, although this is mainly by running 'stateless' and tossing all accumulable data to the GC.

22:01

adlai: if yer actually below 1G, how is 2 'challenge'..?

22:02

*

adlai has always supposed that the 'sufficiently smart tradebot' does some sort of exponential-weighted-accumulation to keep its memory consumption constant-space.

22:03

well, that is my range when using Clozure CL, that has more compact representations for ~everything than SBCL ; back when I used SBCL for this, I routinely had the "toss accumulables to GC" preempted by a segfault killing the whole process.

22:04

consing left & right, eh

22:04

the lame and lazy solution is simply to not accumulate, run (sb-ext:gc) often enough, etc.; much easier than solving hard problems

22:04

yes, parsing JSON websockets conses up a storm

22:04

fortunately both SBCL and CCL have generational collectors

22:05

there are a few places where I sneak in a (rplacd (nthcdr ...) ()), tho

22:06

that is CL's (defun free-tail (linked-list) ..)!

22:07

*

asciilifeform wonders whether sbcl supports pools a la ada

22:07

y'know, 'all conses in the following (progn.. ) to come from this-here array, and will die when it dies)

22:09

that is where, e.g. hash-table with keys of type (vector (3) fixnum) reuses the vectors?

22:09

oblig. ada example of subj.

22:09

adlai: more general.

22:09

*

adlai is probably thinking of something different... yes

22:09

iirc what you described is occasionally called "hash consing" , thus my confusion.

22:10

adlai: it's called 'from this point here to that point there all conses will come from this-here array and when it's full, nomoar consing, and when gc'd, they all get gc'd'

22:11

the ada item, for the curious.

22:15

this seems to not specify what form of failure occurs when reaching the point of 'nomoar consing'

22:15

eggog condition

22:16

I guess this is actually quite a reasonable 'static parallel' to the hash consing idea

22:16

in a hash consing situation, you get colliding objects, instead of running out of space.

22:17

it is only sane when the objects are immutable, and your usecase has multiple references to a bunch of points that are sparse within a much larger set.

22:17

cache

22:18

*

adlai thinks of caching as a much more complex bunch of problems, all dealing with the cases where the data is partially mutable, and thus the cached copies 'expire'

22:19

speaking of expiration, I should eat before the food starts to wake up and plan how to eat me.

22:19

laters then.

22:20