Experimental IRC log happs-2008-08-15

Available formats: content-negotiated html turtle (see SIOC for the vocabulary)

Back to channel and daily index: content-negotiated html turtle

These logs are provided as an experiment in indexing discussions using IRCHub.py, Irc2RDF.hs, and SIOC.

20:27:38<perspectivet>Does anyone have a rough idea +/- 1 month before the multi-master replication stuff goes out in a release?
20:30:21<stepcut>perspectivet: some stuff is available now
20:30:39<stepcut>perspectivet: lemmih gets back from vacation soon, so work should pick up again soon
20:30:47<stepcut>that is all I know
20:31:18<perspectivet>just curious. I helped out a bit on the hspread development at the end of last year and I was wondering how things are going.
20:32:24<stepcut>cool
20:32:25<perspectivet>there have been some nice blog postings at n-heptane labs that make it look like things are starting to come together. Last piece of the puzzle is the multi-master stuff, in my mind.
20:32:48<perspectivet>or for my purposes, rather
20:32:57<stepcut>yeah, I just implemented a simple IRC bot using the existing multimaster stuff, but I don't really understand how some of it works
20:33:16<stepcut>I mostly just changed two lines and started spread. starting spread was the hard part
20:33:31<perspectivet>I saw that in the tuukka irc logs
20:33:31<perspectivet>nice work
20:34:05<stepcut>ACTION wrotes the blogs on n-heptane too, btw
20:34:27<perspectivet>n-heptane is you? cool.
20:34:29<stepcut>yeah
20:34:40<perspectivet>loved the asterisk one
20:35:01<stepcut>I am currently side-tracked on actually development a project using HAppS, but the next step is to try to integrate HSP and formlets.
20:35:17<perspectivet>so you're using happs in production ATM?
20:35:19<stepcut>HAppS-State is probably my favorite haskell library ;)
20:35:41<stepcut>perspectivet: only very limited (two users for a custom report generation tool)
20:35:50<stepcut>perspectivet: but, working on something bigger
20:36:03<perspectivet>stepcut: how's that going, from the standpoint of:
20:36:19<perspectivet>how much is the current state of happs getting in your way, or not
20:36:57<stepcut>HAppS has been pretty good to me. I added support to do on-the-fly (HTML) validation, but I don't think that has been merged upstream yet.
20:37:12<stepcut>HSP had a number of minor bugs, but I got those all cleaned up.
20:38:05<stepcut>the happs-hsp-template dynamic plugin reloading code is a little broken I think (I don't think it regenerates the dependency list after the first compile), but works well enough. Supposedly hsplugins is deprecated -- but I have not figured out what you are supposed to use instead
20:38:30<stepcut>also, OS X, etc, do not support inotify -- so I happs-hsp-template could use a little work there
20:39:12<stepcut>I have started to split the inotify+dynamic code reloading stuff into a standalone library, but it is on hold for this week and until I figure out how to replace hsplugins
20:40:05<perspectivet>sounds like things are coming together.
20:40:13<stepcut>happs seems to run for long periods of time with out crashing or eating RAM, so that is good
20:40:42<perspectivet>I've got a small basic dynamic site project coming up that I'd like to try building in happs
20:40:43<stepcut>hsplugins adds 12MB to the binary size -- that is not so good
20:41:00<stepcut>what OS ?
20:41:06<perspectivet>whatever I want
20:41:26<perspectivet>But it'll be debian GNU/Linux
20:42:26<stepcut>I have everything you need to build happs+hsp applications debianized and built for hardy if you are interested. But it will likely void your warrantee since Ihad to patch the compiler :-/
20:42:57<perspectivet>hehe, i didn't there was a warranty
20:43:05<stepcut>and the happs/hsp code contains some of my own patches (though, I think it is just the validation code at the moment)
20:44:31<perspectivet>cool, I'd be interested in taking a look at that stuff. I won't be working on it until middle of next week, but if you're around I'd love to give it a spin
20:44:40<stepcut>ok
20:45:27<stepcut>the HAppS API has been a lot more stable (in the last 6-8 months) than it was before. Lemmih said that much of the API 0.9 should be stable now.
20:45:48<perspectivet>that's good
20:56:15<perspectivet>stepcut: you're running your sites off ubuntu?
21:00:46<stepcut>perspectivet: for the time being
21:01:20<perspectivet>how are they doing in the security dept these days?
21:01:54<stepcut>perspectivet: no idea -- but since the sites only run SSH and HAppS, I am not too worried -- expect that debian screwed up ssh :(
21:03:15<perspectivet>what did they do?
21:03:39<perspectivet>I haven't really been following debian news over the last 8 months
21:04:26<stepcut>perspectivet: screwed up the random number generator in the ssh-key gerenator, resulting in a very small number of possible keys
21:04:35<stepcut>http://www.ubuntu.com/usn/usn-612-1
21:04:41<perspectivet>ah, right, I did read about that.
21:04:56<stepcut>it was a big racket
21:20:56<stepcut>ultimately, I hope to not even need SSH :)
21:25:18<perspectivet>ah, yes, the dream of the ages
21:26:11<stepcut>that's what we need multimaster for ;)
21:28:36<perspectivet>you mean, just modify your state on a local version of state that's part of the cluster and then voila?
21:29:29<stepcut>or just use the web interface
21:30:32<perspectivet>but then you'd still be using ssl for "secure" communications
21:31:09<stepcut>right, but that is true for intraserver communications as well
21:31:18<stepcut>however, it is one less app that is running
21:31:36<perspectivet>and one less hole to plug
21:31:46<perspectivet>and/or watch
21:31:50<stepcut>mostly, I hope to treat each node as a black box.
21:32:22<stepcut>for example, to deploy an upgrade, create a new ISO with the new version, fire it up on a fresh EC2 node, and then shutdown the old versions.
21:32:57<stepcut>for clusters of any interesting size, I don't think you can be afford to be ssh-ing into individual machines to fix problems
21:33:28<perspectivet>yes, very true
21:33:53<stepcut>if SSH is an important part of debugging the application or fixing things, I think that can only be trouble
21:34:27<perspectivet>agreed
21:34:37<perspectivet>you've gone wrong somewhere
21:34:43<stepcut>during development that is fine, but for deployed applications it can't be
21:34:57<perspectivet>yes
21:35:16<stepcut>I am hoping coyotos gets up to speed soon :)
21:35:27<perspectivet>state migration from version to version needs to be automagic or at the very least easy.
21:35:33<perspectivet>coyotos?
21:35:50<stepcut>the commercial successor to EROS and KeyKOS
21:36:10<stepcut>a very secure microkernel OS
21:36:16<perspectivet>interesting
21:36:23<stepcut>uses capabilities based security instead of ACLs
21:36:36<perspectivet>how's the ghc -> coyotos port going ;)
21:37:02<stepcut>well, GHC is hopefully getting easier to port with the backend rewrite, and coyotos is hopefully getting useable :)
21:38:01<stepcut>the trusted kernel of coyotos may someday be implemented in BitC and formally proven to be correct.
21:38:15<stepcut>but, currently it is just implemented in C :-/
21:38:51<perspectivet>yeah, saw that.
21:39:00<perspectivet>interesting proposition
21:39:37<stepcut>anyway, Jonathon Shapiro is a pretty smart technically and on the business side, and has been hacking on capabilities based systems for a long time now
21:40:21<perspectivet>have you ever looked at CAL?
21:40:33<stepcut>this is the 3rd system he worked on (the first someone else wrote that he learned from, the second was part of a thesis project, and this is the commercial version)
21:40:47<perspectivet>the Haskell-like language for the jvm
21:41:01<stepcut>briefly
21:41:04<stepcut>eating, bbiab
21:41:12<perspectivet>stepcut: coyotos sounds interesting
21:44:53<stepcut>yeah

Back to channel and daily index: content-negotiated html turtle