#ossasepia Logs for 20 Aug 2019

April 19th, 2020
diana_alt: ugh, no deedbot [14:44]
asciilifeform: apparently not [14:45]
diana_alt: asciilifeform: 4 runs of curl http://54.209.217.34/log/trilema/2018-04-18#1802801 > /dev/null give http://p.bvulpes.com/pastes/1llzt/?raw=true [14:49]
asciilifeform: ty diana_alt ! [14:50]
asciilifeform: diana_alt: imho thus far errything points to the ~pipe~ [14:52]
asciilifeform: given e.g. http://logs.nosuchlabs.com/log/trilema/2019-08-19#1929698 [14:52]
snsabot: Logged on 2019-08-19 16:38:26 asciilifeform: ... from here eats 0.268s. [14:52]
diana_alt: BingoBoingo: do you mind waiting until the end of this week for the payment of the Pizarro invoice for my rk? [14:53]
diana_alt: asciilifeform: it would seem so, yes. [14:54]
asciilifeform: diana_alt: btw what sorta load time do you get for 'heavy page' from phf's log [14:55]
asciilifeform: ? [14:55]
asciilifeform: ( his also iirc in usa ) [14:55]
BingoBoingo: diana_alt: No problem at all [14:56]
diana_alt: BingoBoingo: thanks! [14:58]
diana_alt: asciilifeform: hm, a bit longer if anything: real is around 1.5s on 4 runs [14:58]
diana_alt: same precise page [14:59]
asciilifeform: diana_alt: the mirror is on largest spam isp in usa (lulazon) and actually can be moved between a dozen or so locations , we can experiment later if want [15:00]
diana_alt: specifically: http://p.bvulpes.com/pastes/su9en/?raw=true [15:00]
asciilifeform: ( they have a thing where you can pick in what city ) [15:00]
asciilifeform: atm i vaguely suspect that tcp on piz is slowed by an inept wiretap somewhere. [15:01]
asciilifeform: diana_alt: my log pages are slightly heavier than phf's, he had 'minimized' css , some scheme for cutting out unprinted whitespace also [15:02]
asciilifeform: it's not a huge diff tho iirc [15:03]
diana_alt: no, not a huge difference anyway; tbh for the moment I can't see much more info to be gotten out of further simple timing tests (or maybe I just don't see it) [15:08]
asciilifeform: diana_alt: thus far seems to me that we're seeing 100\% pipe effect [15:09]
asciilifeform: ( as confirmed by lobbes , my logtron generates the page in approx same time it takes to move the mouse button. can't say re phf's , likely same ) [15:09]
asciilifeform: i.e. no amt of massage of the proggy is going to make a perceptible diff re load time. [15:10]
diana_alt: I agree it looks 100\% pipe effect [15:10]
asciilifeform: imho what's needed, is http://logs.nosuchlabs.com/log/trilema/2019-08-19#1929756 . [15:11]
snsabot: Logged on 2019-08-19 21:52:50 asciilifeform: rly what we oughta have is 7+ ~active~ bots/logotrons on 7 continents, then no one will have to grumble 'wai slow' [15:11]
diana_alt: asciilifeform: do you recommend just mirroring them on heathen vms? [15:11]
asciilifeform: it poses some difficulty; e.g. the 1 i used, is in fact banned from fleanode (unless sslized) on acct of spam [15:12]
asciilifeform: but proggy is lightweight enuff that it in fact runs a++ in cheapo pseudohosters [15:12]
asciilifeform: (observe that rubbish-mirror has no bot, it is synced via db dump / cron ) [15:13]
asciilifeform: seems that any host service which offers traditional python2, will run the thing. [15:14]
diana_alt: looking at it from the other side, why not just mirror them in the cheapest way too; supposedly there still are some that are not banned on acct of spam either; [15:14]
asciilifeform: it took me just slightly over half hour to set up the example mirror. [15:14]
asciilifeform: ( all of the necessary instructions, i think are in the readme ) [15:15]
asciilifeform: note that the bot doesn't need to be voiced in order to log. [15:16]
diana_alt: asciilifeform: I still tend to do a *2 for myself on all your timings; but in any case, it's still on my list for when I get back to my desk [15:16]
asciilifeform: diana_alt: the moar, the merrier. [15:16]
diana_alt: asciilifeform: maybe add those cron/instructions as a patch though? [15:18]
asciilifeform: diana_alt: it's defo going in next patch. [15:18]
diana_alt: i.e. add to the readme or something so one doesn't need to dig in the logs [15:18]
diana_alt: great, thanks! [15:18]
asciilifeform: imho mirrors synced via cron aint especially useful tho [15:18]
asciilifeform: as they don't help to bridge gaps in log in the event the bot (of which atm there is only 1, of my type) is bumped [15:18]
diana_alt: sure; that doesn't mean they don't help in any other way though [15:19]
asciilifeform: a proper logtron oughta have own bot, going in realtime [15:19]
diana_alt: sure [15:19]
asciilifeform: the lulazon mirror, i set up specifically to probe q of 'is proggy/web stack slow' [15:19]
asciilifeform: as soon as someone sets up proper mirror (i.e. with working bot), we can start on the q of how to auto-sync. [15:22]
asciilifeform: ( it prolly has not escaped diana_alt's notice, that in so doing we're more or less rebuilding the relevant pieces irc ) [15:23]
asciilifeform: *of irc [15:23]
asciilifeform: btw re the cron : the instructions posted in log earlier do not in fact result in a working auto-cronsync ; turns out that it is necessary to run 'init_db' script prior to importing a dump, or postgres barfs . this results in 5 or so min. down time (i.e. machine shows empty log) [15:26]
asciilifeform: i have not conceived of any simple pill against this, of yet. [15:26]
asciilifeform: imho such mirroring is not all that useful, so did not put serious effort into any such thing as e.g. 'trim the dump so doesn't conflict with existing' [15:27]
asciilifeform: any auto-sync really oughta use the 'raw-export' knob (implemented -- emitter, but no receiver of yet) [15:31]
diana_alt: yes, a proper separate working bot (ideally many of them) is the real need; this sort of mirrors are lowest-effort and a rather different thing, can't see any reason to spend a lot of time on them either [15:31]
asciilifeform: diana_alt: http://logs.nosuchlabs.com/log-raw/ossasepia?istart=1000404&iend=1000461 << example [15:33]
asciilifeform: diana_alt: imho errybody in l1 oughta have a working bot. (dun even have to be copy of mine, but oughta emit same 'raw' format ideally) [15:37]
diana_alt: shrysr: those copy/paste bits on your blog - what's exactly their role/your idea with them? they don't quite make posts for sure; are you trying to build a sort of reference list there or what? [15:37]
diana_alt: asciilifeform: I think I even said as much before, yes; it might have been as a comment on trilema though, can't seem to find it quickly right now. [15:41]
diana_alt: the above re every lord should run a bot as an even more basic requisite than running a node [15:42]
diana_alt: there, found it, last line in http://trilema.com/2019/trilema-goes-dark/#comment-130485 [15:44]
shrysr: I replied to your comment... they are not posts. It was meant to be a reference... and I guess I liked the story quite a bit. [22:45]
trinque: funny thing is, looking back on it... it was bad lisp! [22:46]
trinque: the class and method naming at the very least, written like some java dork [22:46]