Overview and Logs for the Dev Meeting Held on 2016-03-19

Postato da: dEBRUYNE / fluffypony


<dEBRUYNE> dev meeting in 5 min, FYI
<hyc> dingdong
<gingeropolous> hello
<dEBRUYNE> fluffypony, smooth, othe, ArticMine, luigiw, NoodleDoodle, tewinget, moneromooo
<floofypony> there we go
<dEBRUYNE> did I forget anyone?
<tewinget> oh, hello
<luigi> is warptangent around?
<ArticMine> Hello
<hyc> he's been fighting a flu last we heard
<moneromooo> hi
<NoodleDoodle> Hello. I'm here but I'm fighting the apocalypse.
<NoodleDoodle> of flus.
<luigi> keep doing it
<luigi> wait you're alive that's good to hear
<dEBRUYNE> fluffypony seems ded
<fluffypony> sorry
<fluffypony> was eating
<fluffypony> welcome everyone
<fluffypony> so as you know we pushed out 0.9.2
<fluffypony> however, there are some nagging issues from the ReadTXN work
<fluffypony> hyc has nailed a major one as of a few hours ago
<fluffypony> so we'll probably do a point release on Monday or so
<fluffypony> also that means that the way we use LMDB has changed a bit
<fluffypony> hyc can you tell us briefly how we should wrapping access to LMDB, both read and write operations?
<hyc> Are you talking about the CRITICAL_REGION stuff?
<fluffypony> yes, and the cursors vs. txns stuff
<hyc> ok, the critical_region stuff actually is not a change at all.
<hyc> basically, when you're setting up to do a write, you need exclusive access to the DB
<hyc> this appears to have been a long-standing bug, unrelated to the readtxn changes
<hyc> so as for reads - there is now a long-lived read txn per thread
<hyc> and a set of read cursors to go with each
<hyc> the TXN_PREFIX_RDONLY macro sets that up in a particular function, grabbing the thread-local-storage for it
<hyc> and RCURSOR(dbname) sets up a read cursor for a particular DB
<hyc> these are analogous to the CURSOR(dbname) macro for getting a write cursor to a DB
<hyc> the point of all this is to avoid a bunch of malloc/free/seek when accessing a DB
<hyc> the old code was allocating a readtxn and cursors inside each function
<hyc> likewise for writes
<hyc> by reusing the same cursors acros a set oof functions we get a pretty good performance gain
<hyc> ok?
<fluffypony> neat
<fluffypony> also on the topic of stuff-hyc-did-lately
<fluffypony> if anyone missed it, we now have a win environment guide up on forum.getmonero.org
<dEBRUYNE> ^ https//forum.getmonero.org/5/support//building-monero-v0-9-2-on-winMonero
<fluffypony> so that should get us all on the same page with testing etc.
<hyc> and one success story replied to it already ;)
<fluffypony> we've also dropped support for BDB as the default database, and switched to LMDB as the default
<fluffypony> including on -bit and ARM
<fluffypony> BDB will remain supported for the moment, primarily as a mechanism for contributors to understand how to build out DB support
<fluffypony> krongle)
<fluffypony> shew we have the entire xmr.to team here today, that's awesome
<binaryFate> fluffypony good memory P
<fluffypony> we shared a podcast together, binaryFate -P
<krongle> yes - impressive nick-name recollection
<fluffypony> hah hah
<fluffypony> while we have you guys here, are you guys doing anything cool you want to talk about?
<binaryFate> we're doing many cool things, but nothing we can talk about at this stage
<fluffypony> hah hah
<fluffypony> it does lead to an interesting point of conversation
<binaryFate> seriously considering btc -> xmr direction
<fluffypony> plugins
<iam6yearsold> If NobleSir or xmr.to team could talk more about xmr.to integration at MiniNero that would be great…. also are 2 way conversions coming to xmr.to soon?
<fluffypony> iam6yearsold Shen's offline at the moment, I'll ask him to update the Reddit thread with some info )
<fluffypony> re plugins, we've spoken briefly about options for the GUI
<dEBRUYNE> iam6yearsold There is a bit of info here -> https//imgur.com/a/HZL7k
<binaryFate> iam6yearsold for MiniNero integration you'd have to see with NobleSir. The API doc is at http//xmrto-api.readthedocs.org/en/latest/
<fluffypony> but I guess we could have "plugins", of a sort, that add functionality
<fluffypony> like xmr.to or shapeshift integration right in wallet2 / wallet2_api
<dEBRUYNE> I think we should be fairly strict on which plugins to allow
<binaryFate> fluffypony we wanted to discuss that plugin integration soon in fact )
<arnuschky> we're quite interested to all secondary questions related to plugins
<fluffypony> I guess the major question becomes
<arnuschky> so plugin repository/db, packaging, distribution etc
<fluffypony> do we allow "permissionless" plugin development, or do we just have a central repo that we git submodule in?
<ArticMine> The main question I see with plugins is trust
<fluffypony> ArticMine exactly
<arnuschky> yes. It puts quite a bit responsibility on the dev team
<gingeropolous> well no ones going to trust anything that doesn;t come from core
<tewinget> <fluffypony> we shared a podcast together, binaryFate -P <– wasn't I there for that?
<fluffypony> considering the recent Google Chrome Bitcoin stealing malware I think that premissionless plugins are dangerous
<gingeropolous> as we've seen with 3rd party GUIs
<luigi> you obviously can't stop permissionless dev, you can discourage users from trusting it I guess
<hyc> we can start signing binaries, ohjoy
<dEBRUYNE> I would prefer the latter
<fluffypony> luigi I mean permissionless within core
<luigi> oh
<luigi> I think no
<arnuschky> it's related to how plugins are written. If it's binary blobs, it's a) hard to build, distribute etc, and b) hard to examine
<binaryFate> fluffypony I'd see both possible all together. Unpermissioned scale and central repo for a selected subset would ease experience and trust for user
<arnuschky> so if plugins are interpreted (eg python), things become a whole lot easier
<iam6yearsold> for the record I hated the twittter plugin idea I saw a while back
<ArticMine> My take permissionless has to be allowed. The end user has to be made aware who is signing and if to trust the plugin
<fluffypony> well the Electrum model works well
<moneromooo> I agree with ArticMine
<arnuschky> (inspection in case of central repo, but also self-distribution by plugin devs)
<fluffypony> ThomasV will merge basically any plugin, as long as it's not malicious
<fluffypony> and plugins are part of the core code, effectively just in a subfolder
<fluffypony> I think it's a testament to Electrum that they haven't had a malicious plugin ever
<arnuschky> how do they deal with the upgrade/maintenance workload? Or do they just disable broken plugins?
<dEBRUYNE> Is there a way to make a subfolder in the subfolder? i.e. (1) signed and approved by core-team, (2) optional
<fluffypony> arnuschky yeah they just disable broken plugins, and eventually they get deprecated out
<ArticMine> We should allow self distribution with appropriate warning
<fluffypony> ArticMine anyone can compile their own build, which would be the same thing
<arnuschky> are you planning for compiled plugins or interpreted ones? that's quite a differens IMHO
<fluffypony> arnuschky so
<arnuschky> self-distribution is a mess for compiled ones…
<fluffypony> I was thinking we have a repo, say it's called monero-plugins
<arnuschky> audit as well
<fluffypony> and then anyone can PR to that repo
<fluffypony> and that repo is pulled into the main Monero source as a git submodule
<fluffypony> there are two advantages to this
<fluffypony> 1. as it gets bigger and harder to deal with, we can step back and other known members of the community can manage that repo
<fluffypony> 2. if we come up with a standard set of functionality hooks, then other projects can pull that same repo in
<fluffypony> eg. jwinterm's lightwallet
<fluffypony> also it means that the compiled Monero binaries have those plugins baked in, and you can't add extra plugins post-compile
<fluffypony> so no need to deal with interpreted code and securing that on disk and in memory
<hyc> baked in means no dynamic loading?
<tewinget> <fluffypony> so no need to deal with interpreted code and securing that on disk <– if securing an interpreted plugin on disk became an issue, securing the binary would be an issue anyway, so I don't know of that bit matters
<binaryFate> Sounds all good to us. If distribution is done through official channels it's great.
<fluffypony> hyc yes
<fluffypony> no dynamic loading
<hyc> cool
<fluffypony> tewinget I mean we can't secure it in the path from random-site-on-the-web down to random-download-folder and eventually into secure-disk-location
<arnuschky> fluffypony that would be really great.
<fluffypony> ok - I think next steps would be to consider some of the hooks we need to provide to add functionality
<fluffypony> we can use the Monero wikia as a collaboration area for that
<ArticMine> It is a good balance for plugins
<fluffypony> and then we'll just update the Monero Slack
<arnuschky> well securing the plugins can always happen by signature - no matter if interpreted or binary
<fluffypony> I'm kidding, we don't have a Slack
<fluffypony> we're not that cool
<aerbax> Would these plugins allow for interpreted languages like Lua or Python?
<arnuschky> :)
<fluffypony> aerbax I don't see why not, individual CMake files in each plugin folder that allow it to produce a library fix everything
<binaryFate> We could financially support development of plugin architecture if xmr.to is the first instanciation of those plugins.
<fluffypony> if it spits out a .so / .a / .dll with the right hooks then it's fine
<tewinget> fluffypony> tewinget I mean we can't secure it in the path from random-site-on-the-web down to random-download-folder and eventually into secure-disk-location <– and yet, we provide binary downloads…so "random site on the web" could be managed the same as said binary downloads
<tewinget> just like any random site on the web can offer binaries for download and we can't secure that either
<tewinget> caveat emptor has to come into play at some point, I think
<fluffypony> the binaries present a single attack surface, and there's GPG-signed hashes
<fluffypony> if we have to do GPG-signed hashes for a bunch of .py files I think I'll go mad -P
<tewinget> I'm not saying I think it should be one way or another, I'm merely pointing out flaws in your argument P
<fluffypony> fair enough
<fluffypony> binaryFate I think the stumbling block will be that somebody needs to champion this and run with it, and I won't be able to lead it due to travelling in a week
<ArticMine> As long as people can compile their own version with non permissioned plugins this can work
<luigi> they can always do that
<fluffypony> yep
<fluffypony> and in fact that would be the testing model
<luigi> we're not apple )
<fluffypony> fork the repo, and build a binary to test your new plugin
<iam6yearsold> asking noobs to compile will limit adoption
<ArticMine> luigi Exactly
<fluffypony> iam6yearsold why would noobs be writing their own plugins?
<gingeropolous> for security
<dEBRUYNE> lol gingeropolous
<fluffypony> lol
<arnuschky> fluffypony championing the first plugin or the whole infrastructure?
<tewinget> What about a curated repo of plugins (as suggested) but with those plugins written in python/lua? Someone modifying the python/lua on a target's disk is the same as someone modifying the binary anyway, and python/lua plugins would be far easier to develop and audit I think
<fluffypony> arnuschky championing the design, I guess
<arnuschky> tewinget I would prefer that.
<iam6yearsold> how about 1 version with binaries and gpg sig and no plugins? caveat emptor for the rest
<arnuschky> mostly due to auditing, and there's no build/distribution mess attached
<fluffypony> I would prefer we remain language agnostic
<fluffypony> iam6yearsold that's what we already have
<tewinget> fluffypony language-agnostic is fine, but…well, how do you imagine that will work out?
<iam6yearsold> thanks pony. I like the current situation then
<fluffypony> tewinget read up
<arnuschky> ah even language agnostic. I thought up to now it's supposed to be a C++ only API
<tewinget> as in, how do we become language-agnostic so that we can remain so?
<fluffypony> [] <aerbax> Would these plugins allow for interpreted languages like Lua or Python?
<fluffypony> [] <fluffypony> aerbax I don't see why not, individual CMake files in each plugin folder that allow it to produce a library fix everything
<fluffypony> ^^
* smooth is here
<fluffypony> also what if a plugin wants to call a function in the core crypto library, for instance?
<arnuschky> design-wise, that's sounds like a nightmare, no?
<moneromooo> Oh, so linked directly ? I kinda assumed it was gointg to be RPC based.
<fluffypony> ok well I think we're getting into an implementation discussion that's outside of the scope of this meeting
<arnuschky> I mean, if you don't have a small and defined API, every bigger change in the wallet will break plugins
<arnuschky> true )
<fluffypony> after the dev meeting we can continue this conversation if you guys want
<fluffypony> but let's first circle back around
<luigi> this deserves some kind of design thread like ringct imo
<moneromooo> Oh, link ?
<fluffypony> moneromooo: "this deserves"
<fluffypony> so nothing yet
<moneromooo> "like ringct"
<fluffypony> oh
<fluffypony> I see what you were asking
<luigi> oh
<moneromooo> Oh
<fluffypony> OH
<luigi> "like ringct is supposed to get"
<moneromooo> Fair enough.
<fluffypony> so basically this is all luigi's fault
<luigi> warp was gonna go it!@
<gingeropolous> its true. i mis-called out luigi on that one
<fluffypony> warptangent is off sick at the moment
<luigi> yes
<luigi> so I blame that
<fluffypony> I blame Canada
<fluffypony> ok back on track
<fluffypony> since the last meeting the bulk of the PRs have been bug fixes
<fluffypony> and changes to the way we access the DB as discussed at the beginning
<fluffypony> we also had a huge discussion about how to handle mixins below the minimum in the RPC call
<fluffypony> which was then implemented in #715
<fluffypony> I'm also going to try to personally spend some time on the text that users see, things like the level 0 logging output and the CLI flag help
<luigi> oh I was gonna do that
<luigi> but then I didn't
<fluffypony> luigi we can do it together
<luigi> awwww
<fluffypony> I can show you the world, shining shimmering splendid
<gingeropolous> take you wonder by wonder
<fluffypony> lol
<fluffypony> also #728 was a little contentious
<fluffypony> so we created a company called Mixinstream that has hired all the contributors
<palexander> heh heh
<fluffypony> and gingeropolous has launched Monero Classic
<gingeropolous> ( sorry )
<fluffypony> -P
<fluffypony> ok so #728 is Ilya's work as part of the GUI effort
<dEBRUYNE> Can I launch unlimited?
<fluffypony> he was struggling with wallet2, and decided to break it out into something more logical and usable
<fluffypony> (to him at any rate)
<ArticMine> What makes it contentious?
<fluffypony> ArticMine I'll get to that now
<fluffypony> he's unintuitively called it wallet2_api, which is a little confusing
<fluffypony> but basically a lot of it is a wallet2_api call which then does little additional logic, and mainly just passes stuff back to wallet2
<fluffypony> and there's a lot of DRY-violating code because of it
<fluffypony> obviously there was some push back, not to prevent merging it
<fluffypony> but more to understand the thought process
<moneromooo> Define DRY ?
<iam6yearsold> DRY violating scares the shit out of me
<gingeropolous> https//en.wikipedia.org/wiki/Don%t_repeat_yourself
<gingeropolous> maybe
<fluffypony> yes
<fluffypony> iam6yearsold DRY violations are just where you have a piece of code in two places
<fluffypony> so any changes have to happen in both
<fluffypony> we can treat the DRY-violating code as a temporary issue, though
<iam6yearsold> two places = more opportunity for error
<fluffypony> as we're going to wait until Ilya is done with it
<ArticMine> Which makes maintenance much harder
<fluffypony> and then we'll either drop wallet2 and replace it with the new wallet API
<fluffypony> (ie. replace the simplewallet calls as well)
<fluffypony> or if it's just a pointless layer we'll have to go the opposite route
<fluffypony> and change his Qt callers to use wallet2
<fluffypony> as it stands it's kinda undecided and we'll have to see how Ilya goes
<ArticMine> Is Ilya aware of the concern?
<fluffypony> ArticMine yes, we had some discussion on the PR, and othe has also spoken to him afaik
<fluffypony> he was responsive on the PR comments, but this isn't Bitcoin
<fluffypony> we don't ACK NACK utACK for years before merging somethingm
<fluffypony> aintnobodygottimeforthat.gif
<luigi> utNACK
<fluffypony> luigi #networknerd
<moneromooo> utACK was not a typo ?
<luigi> no
<luigi> means untested
<luigi> conceptACK or similar
<fluffypony> yeah
<fluffypony> moneromooo https//lists.linuxfoundation.org/pipermail/bitcoin-dev/-December/71.html
<fluffypony> if you're interested
<hyc> crap
<fluffypony> LOL
<fluffypony> PasteGate 2.0
<gingeropolous> internets
<othe> ur pasting skills suck
<dEBRUYNE> Hahah
<fluffypony> othe pasting is a scam
<hyc> that's how I write all my patches
<fluffypony> I just copy-and-paste code from StackExchange
<gingeropolous> thats my job
<fluffypony> heh
<fluffypony> ok so anyone who can reproduce the 0.9.2 segfault please try latest master
<fluffypony> and if you're still segfaulting let us know
<fluffypony> else we're going to do a point release on Monday
<fluffypony> 0.9.3, I guess?
<luigi> hrm
<fluffypony> or
<luigi> we're gonna run out of numbers at this rate
<fluffypony> yeah we are
<luigi> oh wait
<hyc> 0921
<luigi> we have 0.10
<luigi> nevermind
<iam6yearsold> will there be multiple devs in IRC at time of hard fork this week just in case? I see a few pools still on old cold and probably a few users too
<fluffypony> yes we just do a Bitcoin
<moneromooo> No chance, there's an infinity of those.
<fluffypony> 0.11
<fluffypony> iam6yearsold yes, and we've reached out to as many of them as we can
<luigi> is 0.10 supposed to be for next hard fork?
<fluffypony> luigi fork that
<fluffypony> when warptangent is back we'll see how it goes on ringCT
<fluffypony> and make a more concrete decision as to the timing of the next fork
<dEBRUYNE> iam6yearsold The hardfork will approximately take place at 13:00 UTC, so both US and Europe will probably be awake
<dEBRUYNE> and Asia of course
<luigi> everyone will be awake
<fluffypony> even me
<dEBRUYNE> hawaii will probably be asleep
<fluffypony> heh
<dEBRUYNE> fwiw!
<luigi> wat
<Wolf`> lol
<smooth> we should also consider what else we should go in the next major version besides ringct (doesn't need to be discussed now)
<dEBRUYNE> uh I meant UTC btw
<dEBRUYNE> you muricans with AM/PM
<Wolf`> who got drunk and posted about a party in #monero-dev
<luigi> oh
<luigi> then america won't be up
<moneromooo> The db reorg seems like a good candidate.
<luigi> oh well
<fluffypony> smooth agreed
<dEBRUYNE> east coast will right?
<luigi> sure ish
<dEBRUYNE> You better set your alarm luigi
<fluffypony> Surae is also going to be picking up MRL-6 in the summer
<fluffypony> he has some ideas about how to complete that
<dEBRUYNE> MRL-6 is multisig?
<iam6yearsold> I will party hard if fork happens with no drama
<fluffypony> dEBRUYNE: no
<fluffypony> https//github.com/monero-project/research-lab/tree/master/publications/MRL-%-%Difficulty%Adjustment%Algorithms%in%Cryptocurrency%Protocols
<dEBRUYNE> oh cool, thanks
<moneromooo> How do get cmake to tell you the commands it's running ?
<luigi> we have diff, we have db stuff, we have fee stuff
<fluffypony> moneromooo: I normally make VERBOSE=1
<moneromooo> Thanks, I was trying V=1
<luigi> I like my V=2
<fluffypony> ok - any last things to add
<fluffypony> or can we call it?
<luigi> call it

Post tags : Dev Diaries, Monero Core, Cryptography