IRC channel logs

2016-10-14.log

back to list of logs

<dsmith>Ah. Forgot to mention.
<dsmith>I'll be in Boston (well, "Woburn" wherever that is) the 17-19
<dsmith>Would be cool to meet some Guilers in the area
<wingo>o/
<wleslie>hi!
<wingo>ohai
<wleslie>enjoyed your blog post
<wingo>oh tx
<wingo>tx for the early discussion re: glyph's article too
<wingo>the specific case that glyph discusses is even more egregious in a concurrent ml system, as all channel sends are synchronous, and thus a channel-based logging facility would be synchronous too
<wingo>a system with preemption & multi-threading can only work if there is good software culture, whether enforced or simply encouraged by the language
<wingo>well
<wingo>s/good/appropriate/
<wingo>python's software culture is good too of course :)
<wingo>but i mean that programmers in a preemptive environment should be verrrrry wary about unstructured mutation
<wingo>maybe reagents can help improve that case
<wleslie>that's sort of why Armin had originally tried to go with STM, aiming to reduce the cognitive difference that preemption brings to existing single-address-space systems
<wingo>ACTION nod
<wleslie>but yes, if you don't have mutation all over the place, there are better options
***vlad___ is now known as ivladak-work
***ivladak-work is now known as vlad___
***ivladak-work is now known as newnickname
***newnickname is now known as ivladak-work
***ivladak-work is now known as ivladak_
***ivladak_ is now known as ivladak-work
<wingo>ecraven: how is the benchmarking going? :)
<ecraven>wingo: hehe, well, finished a few days ago, and I totally forgot about them!
<ecraven>thinks for reminding me
<wingo>:)
<rekado>ugh, a blogger who is well-known in Germany wrote about the recent Guile security fix. Apparently without understanding the issue.
<wingo>they think perhaps that all guile programs are vulnerable, that would be my guess?
<wingo>i admit i had that impression a bit from the initial announcement
<rekado>yes, as if all Guile processes by default listen on localhost for code to execute
<rekado>the quote is: "The default in Guile has been to expose a port over localhost to which code may be passed. The assumption for this is that only a local user may write to localhost, so it should be safe."
<rekado>that sounds very misleading in isolation
<wingo>yeah. it's a funny thing, you want to explain the fix, you want users to be secure, but you don't want to overstate the issue; and perhaps this time we overstated the issue i guess
<wingo>oh well. it will blow over
<janneke>'Guile? Never heard of that, what a funny name! Better have a look.' ;-)
<wingo>:)
<wingo>i understand it's a funnier name in german
<OrangeShark>hello everyone
<void-pointer>Wait, it's a funnier name in German? I guess I need to learn a lot more German then
<void-pointer>need to anyways since I live in Germany
<janneke>hi OrangeShark!
<janneke>void-pointer: `Guile'
<janneke>guess the possible diphthong gets the german language area confused
<profan>ACTION still voting fo renaming Guile to Geil
<void-pointer>Never did know how people pronounce it
<void-pointer>I've always pronounced it like GwILL
<void-pointer>with the 'w' just barely being there
<wingo>that's not the right way :)
<wingo>right way is guy ell
<void-pointer>got it
<void-pointer>thank you
<rekado>I pronounce it like the German “geil” (~ horny)
<rekado>(same pronunciation as “guy ell”)
<void-pointer>I take it you live or have lived in Germany?
<OrangeShark>How would one live code a web server using guile?
<rekado>void-pointer: yes, I’m living in Berlin now.
<wleslie>ACTION attempts
<wleslie>guttenackt, guillern!
<rekado>:)
<void-pointer>I'm living in Göttingen, though at this moment I am out of the country.
<davexunit>we're on the HN front page https://news.ycombinator.com/item?id=12705740
<davexunit>...due to the vulnerability
<wingo>lol
<janneke>yay!
<wingo>well the comments appear to be good anyway
<wingo>what is happening to the world
<janneke>sanity and tolerance, that will be the end of us
<janneke>tinkering with syntax-case and implementation of quasisyntax/unsyntax
<paroneayea>whew
<paroneayea>glad the comments are good
<davexunit>me too
<paroneayea>I hope people can take it less as "Guile is making this mistake" than "oh man, like nearly all software makes this mistaken assumption / implementation issue"
<paroneayea>HN for once seems to be getting the message.
<davexunit>yeah, I was pleasantly surprised.
<paroneayea>oh http://lwn.net/Articles/703485/
<paroneayea>lwn too
<paroneayea>they did a good job of quoting there
<davexunit>interesting how the HN comments are of higher quality than the LWN comments
<paroneayea>yeah usually I find the reverse is true.
<davexunit>today is bizarro day
<dsmith>Happy Friday, Guilers!!
<civodul>hey, happy Friday dsmith!
<civodul>dsmith: we miss sneek!
<janneke>wondering if sneek misses us too
<civodul>:-)
***karswell` is now known as karswell
<dsmith>Oh he's gone?
<dsmith>Fouey
<dsmith>civodul: Any idea when it went away?
<OrangeShark>dsmith: looks like on Sept 24 according to the logs
<OrangeShark>there was a lot of net splits and stuff that day too
<wingo>1 month!
<dsmith>Ouch
<dsmith>I'll poke the box tonight
<amz3>catonano: I saw your message on guix help
<catonano>yeah :-/
<amz3>catonano: I think there is no hard requirement on guile-next, otherwise said, you can use guile
<amz3>2.0
<catonano>oh
<catonano>ok, I'll try. Thanks !
<amz3>I mean I don't remember using anything particular to 2.2
<amz3>that said it's slows ;)
<catonano>slower than 2.2 ? My dataset is very small now, I won't notice
<paroneayea>guix-guile-next
<catonano>paroneayea: what ?
<amz3>it's password
<amz3>it's a password
<l_zzie>Hi! I can start a repl server with --listen=something. How do I connect a client to it?
<l_zzie>or rather -- what program do I use as a client?
<jmd`>What does the starred version of define-record-type do??
<jmd`>(yes I have read the manual)
<random-nick>jmd`: I think that's a guix thing
<jmd`>Oh
<jmd`>Isn't that a bit unethical of them?
<amz3>there is a lot of define-record-type* around
<random-nick>some people think that a lot of the guix utility functions for different things should be moved to guile proper or at least separated into a library
<amz3>like https support ;)
<davexunit>that's part of gnutls
<davexunit>not guix
<l_zzie>oh, found the manual. nc / telnet works.
<amz3>l_zzie: you can use geiser
<janneke>it seems than unsyntax works quit differently than unquote
<janneke>ACTION is terribly confused
<paroneayea>l_zzie: hi! you probably want to use emacs + guile + geiser
<paroneayea>if you're looking to do live hacking
<paroneayea>l_zzie: I also recommend doing --listen=/tmp/guile-socket and using "M-x geiser-connect-local" to connect to it. See the most recent release for details.
<OrangeShark>hey paroneayea, what is your workflow for live hacking a web service? Did you do that with pubstrate?
<paroneayea>OrangeShark: I did do it for pubstrate. It depends on if you're using one of the new async systems or just the built in http system.
<paroneayea>OrangeShark: there's a quick-and-dirty-solution, which is what I've used for pubstrate so far, because I haven't moved it over to a nicer async solution yet
<OrangeShark>I see
<OrangeShark>what did you do for the built it http server?
<OrangeShark>*built in
<paroneayea>OrangeShark: "spawn-server"; see 6.17.12 "REPL Servers"
<janneke>trying to get syntax-case up...
<janneke>either eval comes too soon: eval: unbound variable: exp
<OrangeShark>ah, so it runs it in another thread
<janneke>or much too soon: eval: unbound variable: process-rule
<janneke>...or too late: syntax-error:urg2 input=:((sc:when #t (display HELLO-SC-WHEN) hello-sc-when))
<OrangeShark>thanks paroneayea :)
<stis>darn folding a weak hash table is allocating memory
<stis>hmm maybe not
<janneke>unsyntax recurses deep to remove all (syntax ...) wrappers?
<stis>Im inside gc hooks and want to fold over a weak key hash table, currently that blocks
<stis>the reason seams to be that the folder takes a gc lock that is already held
<stis>The solution is to have a separate folder for use in gc hooks
<stis>The situation is sad because maintaining a set of objects for advanced gc algorithms needs something like a weak key hash table
***Fuuzetsu_ is now known as Fuuzetsu
***_hanhart_ is now known as _hanhart
<taylan>janneke: unsyntax doesn't do anything; it's just a "token" scanned for by quasisyntax.
<taylan>janneke: and all it does is that the unsyntax'd form is evaluated normally rather than as literal syntax. this means that e.g. further (syntax ...) forms within it will just become syntax. if you know how quasiquote works; it's exactly like that.
<taylan><janneke> it seems than unsyntax works quit differently than unquote
<taylan>that shouldn't be the case, unless I'm very confused right now
<taylan>janneke: to get a feeling for syntax/quasisyntax/unsyntax, use them outside the context of macros/syntax-case. you'll note you can use them exactly like quote/quasiquote/unquote, just constructing syntax objects rather than raw sexprs.
<amz3>I have brillant idea, but I'm too sick to code :(
<amz3>ACTION goto sleep
***masoudd is now known as LjL_
***LjL_ is now known as masoudd