couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF IRC Services" <asf...@wilderness.apache.org>
Subject Summary of IRC meeting in #couchdb-meeting, Wed Feb 20 20:05:02 2013
Date Wed, 20 Feb 2013 20:55:24 GMT
Members present: Noah, eckoit, davisp, kocolosk, kzx, benoitc, daleharvey, wendall911, rnewson,
tilgovi, chewbranca, dch

----------------
Meeting summary:
----------------

1. Preface
  a. benoitc or eckoit to bring up topic of bundling erica in leu of plugin system (Noah,
1)

2. erica

3. topics
  a. benoit working hard on rebar branch (Noah, 3)

4. 1.3.0
  a. noah working on automation scripts for building releases, holding up actual release (Noah,
4)

5. distro packaging, and its future
  a. wendall911 to write up thoughts on state of 3rd part packaging and send to list (Noah,
5)
  b. http://mail-archives.apache.org/mod_mbox/couchdb-dev/201301.mbox/%3CCAL+Y1nvHe9HaF_rPh-wseJ5BzSWkycV46UGYCL7J1fKRnm1ggQ@mail.gmail.com%3E
(dch, 5)

6. roadmap proposal
  a. wendall911 +10'd our roadmap proposal (Noah, 6)
  b. wendall911 really liked what jan____ did over the discussion on floats (Noah, 6)


--------
Actions:
--------
- benoitc or eckoit to bring up topic of bundling erica in leu of plugin system (Noah, 20:23:56)
- wendall911 to write up thoughts on state of 3rd part packaging and send to list (Noah, 20:47:01)

IRC log follows:


# 1. Preface #
20:06:04 [Noah]: okay, i'll just go around people
20:06:11 [Noah]: dch do you have anything you'd like to discuss?
20:07:17 [Noah]: okay, moving on
20:07:25 [Noah]: jan____: do you have anything you'd like to discuss?
20:08:25 [Noah]: okay, moving on
20:08:32 [Noah]: JasonSmith: do you have anything you'd like to discuss?
20:09:25 [Noah]: okay, moving on
20:09:33 [Noah]: jsonified: do you have anything you'd like to discuss?
20:09:47 [Noah]: dch could you do a name check in #couchdb-dev for me?
20:09:55 [Noah]: maybe announce the meeting in dev and the regular channel
20:10:12 [dch]: Noah: no,  nothing to add atm, I'll do the ping thing
20:10:33 [Noah]: okay, moving on
20:10:42 [Noah]: kocolosk: : do you have anything you'd like to discuss?
20:10:56 [kocolosk]: nop
20:11:09 [Noah]: okay thanks!
20:11:17 [Noah]: rnewson: do you have anything you'd like to discuss?
20:12:32 [Noah]: tilgovi: do you have anything you'd like to discuss?
20:12:47 [tilgovi]: nope.
20:12:48 [Noah]: cool thanks
20:13:02 [Noah]: benoitc: do you have anything you'd like to discuss?
20:13:25 [davisp]: We have a meeting channel?
20:13:40 [Noah]: davisp: yep, welcome!
20:14:17 [Noah]: chewbranca: do you have anything you'd like to discuss?
20:14:32 [chewbranca]: nope
20:14:32 [Noah]: thanks
20:14:32 [wendall911]: Comment: Wanted to say that the roadmap proposal looks incredible.
20:14:41 [Noah]: daleharvey: do you have anything you'd like to discuss?
20:15:02 [Noah]: wendall911: ill get to you in a few minutes!
20:15:40 [davisp]: wendall911: where's that at?
20:15:41 [Noah]: drsm79|away: do you have anything you'd like to discuss?
20:15:47 [daleharvey]: nope, good here cheers
20:15:55 [Noah]: thanks daleharvey 
20:16:42 [Noah]: eckoit: do you have anything you'd like to discuss?
20:16:47 [eckoit]: yeah, bundling couchapp (erica impl) with couchdb until we get cpm
20:17:02 [Noah]: cool, our first taker. today is a slow day :)
20:17:03 [eckoit]: heh
20:17:04 [Noah]: what is cpm?
20:17:32 [eckoit]: (a theoretical couch package manager)
20:17:32 [Noah]: ah... couch package mana... right
20:17:40 [Noah]: yeah - i dont think that's a good idea any more
20:17:50 [Noah]: jan____ has convinced me that we should keep couchdb lean
20:17:55 [Noah]: i mean that's the whole reason behind cpm
20:18:04 [Noah]: i don't think we want to be adding more stuff in the mean time?
20:18:17 [Noah]: community efforts should be rallied around the plugin system if possible
20:18:19 [eckoit]: do we want to discuss now, or collect topics?
20:18:25 [Noah]: actually discuss things
20:18:32 [eckoit]: ok
20:18:47 [Noah]: so go ahead
20:18:47 [wendall911]: I think having per db configurations (on the roadmap) would be necessary
as well if incorporating erica
20:19:10 [benoitc]: what is per db config ?
20:19:32 [benoitc]: jusrry just came to my mind that we may habe some kind of meeting today
20:19:40 [eckoit]: benoit has done the work of bundling erica with refuge/rcouch and want
to move that into couch 
20:20:04 [eckoit]: as dealing with ddocs from the file system is pretty standard
20:20:17 [Noah]: benoitc: you dont have a calendar reminder?
20:20:32 [wendall911]: 23. per-db overrides of server-wide settings ... Only asking since
I'd want to ensure that my existing couchapps woudn't break if erica was enabled. Unless that's
not an issue at all.
20:20:34 [benoitc]: Noah: i don't have any calendar
20:20:40 [Noah]: :/
20:21:05 [eckoit]: wendall911: having erica just provides a std way of loading couchapps included
with couch. all other couchapp loaders still would work
20:21:40 [wendall911]: eckoit: ok, then not an issue at all
20:22:19 [benoitc]: wendall911: i think it's othogonal
20:22:19 [Noah]: do we have any #info or #action items?
20:22:40 [benoitc]: thoughthat let remember me i need to add such thing to https://github.com/benoitc/libcouch
20:22:55 [wendall911]: benoitc: ok, that makes sense based on what's current. I have just
passing knowledge of erica
20:22:55 [eckoit]: ok I will wait for the topic to come back
20:22:55 [benoitc]: Noah: maybe discussing the possibility on the ml
20:23:17 [benoitc]: like we did for f*ton
20:23:19 [Noah]: okay
20:23:56 [Noah]: #action benoitc or eckoit to bring up topic of bundling erica in leu of plugin
system


# 2. erica #
20:24:17 [Noah]: this whole ping based model sux and i am abandoning it half way through
20:24:32 [benoitc]: what to you mean?
20:24:40 [Noah]: i was going around people pinging them. its a stupid model for a meeting
20:24:48 [Noah]: okay, i am collecting topics. half way through. heh
20:24:55 [Noah]: not many of us here
20:25:10 [Noah]: so wendall911 eckoit davisp benoitc dch etc - topics?
20:25:12 [Noah]: for me: 1.3.0 release
20:25:25 [dch]: can I help with 1.3.0 in anyway? other than that, nothing new


# 3. topics #
20:25:32 [dch]: meta woah.
20:25:32 [benoitc]: i'm not sure  i have one actually. i can say i'm working hard on having
a rebar branch in couchdb repo
20:25:42 [benoitc]: still working on a pluging for icu blah
20:25:47 [Noah]: #info benoit working hard on rebar branch
20:25:55 [Noah]: any other topics?
20:26:10 [Noah]: vmx?
20:26:10 [wendall911]: I wanted to discuss distro packaging, and it's future
20:26:17 [Noah]: okay, noted
20:26:40 [Noah]: okay, pressing on


# 4. 1.3.0 #
20:26:49 [Noah]: current status: blocking on me
20:26:55 [Noah]: i am working on this every night at the moment
20:27:02 [Noah]: the tests are very painful for me
20:27:02 [Noah]: so i am working on automation
20:27:11 [benoitc]: oh i actually have one : websockets
20:27:18 [Noah]: once i have this finished i should be able to issue one command and have
a release plop out of the other end
20:27:34 [Noah]: unfortunately, because of the tests, the time it takes to do this could be
hours, days even, who knows?
20:27:40 [dch]: Noah: can we help at all? I've not seen the issues you experience.
20:27:47 [Noah]: i will have more details once ive run it successfully
20:27:55 [Noah]: dch: nope - wait for me to post results
20:28:02 [Noah]: i can give you steps to reproduce however
20:28:10 [dch]: for sure,
20:28:10 [Noah]: spin up an xlarge EC2 AMI
20:28:17 [Noah]: do the release procedure
20:28:17 [Noah]: that's it
20:28:25 [benoitc]: i really dislike our test suite
20:28:25 [dch]: debian or ubuntu or sth?
20:28:25 [Noah]: (choose ubuntu)
20:28:41 [Noah]: benoitc: ive given up moaning about it and am shell scripting my way around
it
20:28:47 [benoitc]: actually someone should work on that 
20:28:47 [dch]: interesting. I run these normally in a vmware fusion vm, and never have problems.
20:28:55 [Noah]: my script basically just brute forces the thing
20:28:57 [Noah]: as in runs it over and over
20:28:58 [wendall911]: Noah: rnewson has mentioned doing this quite a bit in the past. I think
that's what they do for package building at cloudant, iirc
20:29:02 [dch]: I saw the screenshot, good move.
20:29:03 [Noah]: only when it passes, does it stop
20:29:11 [Noah]: screenshot?
20:29:18 [Noah]: ah that i posted to jan in #couchdb?
20:29:19 [Noah]: yeah
20:29:34 [benoitc]: Noah: basically i think we should remove all the js tests. anyway that
is another topic. and i've no time to work on it until april
20:29:34 [Noah]: jan wants to move us to a jenkins system so we can just nap a package and
call a vote on it
20:29:40 [rnewson]: (https://github.com/cloudant/packages)
20:29:40 [dch]: if you can send me what you have atm, I will windows-ify it.
20:29:55 [dch]: Noah: I agree, we are not too far off this.
20:30:02 [Noah]: #info noah working on automation scripts for building releases, holding up
actual release
20:30:17 [Noah]: dch ill send you it once its sorta complete
20:30:25 [benoitc]: https://github.com/refuge/rcouch/tree/master/package as well
20:30:27 [Noah]: dch would be interested to see if you can replicate on aws
20:30:40 [dch]: I'll try that tonight.
20:30:56 [Noah]: i also want to do a script that will automate release testing
20:31:02 [Noah]: might increase instances of people chipping in
20:31:11 [Noah]: anyway i am working on this every night atm - will post more when i have
more
20:31:17 [wendall911]: what will it take to kill the futon test suite? Is it a 1.3.0 issue,
or some future change?
20:31:32 [Noah]: wendall911: its not the futon tests that are causinig me problems - its the
cli ones
20:31:40 [rnewson]: are we planning on ever killing it?
20:31:47 [dch]: BTW what about simply removing the actual link right now, for 1.3.0?
20:31:55 [Noah]: actual link?
20:31:56 [wendall911]: Noah: those are all problematic on slow hardware, or VMs in my experience
20:32:03 [dch]: the underlying url can stay, but putting it right there in the futon console
begs people to try it.
20:32:10 [Noah]: wendall911: unfortunately, vms are all i has
20:32:25 [Noah]: dch its not causing us any problems though
20:32:32 [Noah]: dch like, certainly not for 1.3.0
20:32:35 [wendall911]: there were a bunch of fixes for the 1.2.0 release that jan____ put
in to address some of the problems, but I know they still randomly fail for me occasionally
20:32:40 [rnewson]: Noah: I can spend some time with you to smooth this out, or at least find
out why the test suite hates you.
20:32:55 [Noah]: rnewson: once i have this script done i will share it, and the results of
running it and you can see if you can replicate
20:33:02 [rnewson]: thanks.
20:33:17 [Noah]: rnewson: just trying on an ubuntu ami will give you the results though (i
presume, unless they literally hate *only me*)
20:33:25 [rnewson]: *nod*
20:33:32 [Noah]: okay moving on


# 5. distro packaging, and its future #
20:33:40 [rnewson]: my build scripts in the above repo start with a 'knife ec2 create' line.
20:33:41 [dch]: I think it's something personal. Perhaps AWS was a cheese pastie in a former
life
20:33:50 [wendall911]: ok, packaging is changing all over
20:34:11 [wendall911]: I think node.js really started the  move, but PEP 427 was just approved
20:34:17 [Noah]: rnewson: hmm..
20:34:40 [wendall911]: and there is a lot of discussion about doing something similar in couchdb,
app registry or whatnot
20:35:30 [wendall911]: I think we should have something like pep 427 for couchdb for downstream
packagers to understand the logic around the packaging strategies. Old methods clash with
some of the needs for couchdb. Especially for things like js engines, etc.
20:36:08 [eckoit]: wendall911: what is PEP 427?
20:36:15 [dch]: http://www.python.org/dev/peps/pep-0427/
20:36:17 [wendall911]: Distros do a crap job of supporting multiple libraries, so my concern
is that if couchdb has a hard dependency on a spidermonkey version (which it should) that
this will be problematic getting into distros.
20:36:23 [rnewson]: YAPF (yet another packaging format)
20:36:25 [dch]: PEP is a python proposed enhancement
20:36:30 [wendall911]: sorry, for not providing the link
20:36:53 [dch]: 427 is a zip file format for installing python things
20:37:00 [wendall911]: rnewson: yeah, basically that, but couchdb will suffer the same issue
20:37:45 [benoitc]: wendall911: so the rebar branch will allows you to build an erlang release
20:37:47 [dch]: wendall911: my suggestion is to recommend to people re-use the cloudant spidermonkey
packages, and that pretty much fixes all the issues we have for people building from source.
20:37:53 [wendall911]: dch: the implications of 427 is really what I'm getting at, not the
packaging format
20:38:08 [benoitc]: so you can just provides static binaries of couchdb using that
20:38:45 [rnewson]: dch: +1
20:39:00 [Noah]: wendall911: do you have any action items for us?
20:39:00 [wendall911]: dch: but that's the equivalent of telling the distros to fuck off,
and that's not really a very good position
20:39:33 [rnewson]: enh, post-merge the distros are going to be pissed anway.
20:39:38 [wendall911]: I'd like to see couchdb have a stance of cooperation with how screwed
up distro packaging is, and that's the python pep 427 approach
20:39:45 [benoitc]: imo that could be solved by replacing sm by v8 and embed v8 in our code
just like nodejs
20:40:08 [wendall911]: benoitc: but, they'll still strip v8 out and use the system package
20:40:08 [benoitc]: i had started some couchjsv8 but was distracted by more urgent tthings
20:40:15 [dch]: wendall911: the distros still have the option to build from source their packages,
depending on R14A and R15B02, and with a spidermonkey that changes over time.
20:40:23 [benoitc]: wendall911: and this their problem right after
20:40:45 [wendall911]: yeah, I suppose so, but it annoys me that RHEL 5 is still on 1.0.4
20:40:45 [benoitc]: v8 is more conservative and stable across releases anyway
20:40:53 [wendall911]: and that reflects poorly on couchdb, imo
20:40:54 [benoitc]: so it's easier than using sm
20:40:55 [rnewson]: the freedom to use R14A is theirs, as stupid as that is, but we don't
have to help.
20:41:18 [wendall911]: what other projects have done is provide their own package repos, and
tell users that the maintainers are doing a poor job
20:41:26 [wendall911]: nginx for example
20:41:33 [dch]: wendall911: I'm not advocating for badmouthing maintainers.
20:41:45 [Noah]: wendall911: i think we might want to look in to that
20:41:53 [Noah]: wendall911: an apache dep / rpm repo would be very cool
20:41:55 [rnewson]: dch: I have some unkind words for the folks that chose a beta release
of erlang/OTP for the stable release of an OS.
20:42:00 [wendall911]: dch: I'm not advocating that at all, but it's the truth
20:42:09 [Noah]: wendall911: is that something you'd like to help with?
20:42:18 [dch]: the reality is that apache release cycles are in orders of months, and getting
into stable releases of OS distros is on order of years.
20:42:30 [dch]: and this is not going to change in the forseeable future.
20:43:15 [wendall911]: Noah: I'm well versed in packaging, as are several others, but I think
it would be a stronger position to have "official" packages. I thought Wohali was heading
that up
20:43:15 [dch]: rnewson: +1 to a deserved spank for those people. If you're packaging couchdb
you should understand the A in OTP, and be aware of the 64-bit R14B02 booboo. etc etc.
20:43:23 [rnewson]: dch: certainly.
20:43:23 [benoitc]: a lot of projects have their ppa blah. the question is then why we don't
have repos done by voluntars
20:43:30 [Noah]: there's nothing we can do about getting official packages
20:43:38 [benoitc]: and maybe it's because our current build is too much crappy
20:43:38 [Noah]: we cant force other people to do work
20:43:52 [benoitc]: no but we can make their life better
20:43:52 [rnewson]: my meta-point is that an properly packaged erlang release, our intention,
will be self-contained by design.
20:43:55 [Noah]: but we can provide our own package repository
20:44:16 [Noah]: are there any action items here?
20:44:31 [Noah]: does anyone want to put together a proposal, or start a discussion? wendall911?
20:45:00 [rnewson]: I'm +1 on providing deb/rpm/etc ourselves.
20:45:00 [kzx]: (sorry to interject) epel 6 and 5 where couchdb packaged go in won't be able
to move to a new major version of erlang they are stuck at r14 and r12 respectively (Peter
Lemenkov mentains them) that's the policy, and epel 6 presumably will be around for many years,
 i would rather have a single package built from an erlang release that statically includes
as many dependencies as possibly and just goes into /opt/couchdb/ for exam
20:45:02 [wendall911]: Noah: I'll write something up for the list. I'll see if I can get Woholi
on board
20:45:32 [benoitc]: i'm a big -1 on porviding our own packages
20:45:38 [benoitc]: this isn't our role
20:45:38 [Noah]: rnewson: is that something cloudant could help with? you already seem to
be doing stuff around that
20:45:55 [rnewson]: Noah: we don't build packages though, unless forced to.
20:46:00 [wendall911]: benoitc: the current situation harms couchdb, imo
20:46:00 [Noah]: benoitc: it would be the case that someone like rnewson provides them on
his own accord for the convenience of the community
20:46:07 [rnewson]: we make erlang releases and install them.
20:46:23 [wendall911]: there isn't any motivation for distros to ship anything new, so they
don't
20:46:23 [benoitc]: Noah: i prefer to make erlang release
20:46:30 [benoitc]: which will help packager to build their packages anyway
20:46:33 [rnewson]: I have scripts to make bigcouch .deb/.rpm's but I think I'm the only one
that does so.
20:46:37 [benoitc]: it's a lot easier
20:46:37 [wendall911]: benoitc: that makes more sense
20:47:00 [benoitc]: rnewson:  nope :) https://github.com/refuge/rcouch/tree/master/package
but it is building around the release
20:47:01 [Noah]: #action wendall911 to write up thoughts on state of 3rd part packaging and
send to list
20:47:08 [Noah]: can we move on?
20:47:08 [wendall911]: rnewson: I do the same for couchdb every release.
20:47:15 [benoitc]: and I have now some people that will do on debian
20:47:32 [dch]: #info http://mail-archives.apache.org/mod_mbox/couchdb-dev/201301.mbox/%3CCAL+Y1nvHe9HaF_rPh-wseJ5BzSWkycV46UGYCL7J1fKRnm1ggQ@mail.gmail.com%3E
20:47:38 [wendall911]: yeah, we can move on, but I do agree with benoitc, fixing the release
will be the fix


# 6. roadmap proposal #
20:47:52 [Noah]: wendall911: you had some thoughts on this?
20:48:00 [dch]: wendall911: if it makes sense, please continue that thread for packages etc.
20:48:07 [wendall911]: Oh, just wanted to say it's very, very good
20:48:15 [wendall911]: the proposal
20:48:16 [Noah]: which bit specifcially?
20:49:09 [wendall911]: all of it. I think it covers every issue I can think of well.
20:49:30 [wendall911]: I'd +10 if I could
20:50:52 [Noah]: okay.. hehe
20:50:55 [Noah]: well, we're out of topics
20:51:00 [Noah]: anyone got anything to discuss very quickly?
20:51:22 [wendall911]: Just one quick thing
20:51:38 [wendall911]: I really liked what jan____ did over the discussion on floats
20:52:00 [Noah]: #info wendall911 +10'd our roadmap proposal
20:52:10 [Noah]: #info wendall911 really liked what jan____ did over the discussion on floats
20:52:11 [wendall911]: move it to the documentation as an artifact. I think this is missing
sometimes, and very useful information gets buried in the list archive
20:52:15 [Noah]: :)
20:53:16 [wendall911]: that's all, want more of that :)
20:53:41 [Noah]: okay
20:53:48 [Noah]: well i think that's a wrap folks
20:54:00 [dch]: thanks Noah!
20:54:00 [Noah]: np


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message