16:03:24 #startmeeting VPP Weekly 16:03:24 Meeting started Tue Feb 16 16:03:24 2016 UTC. The chair is edwarnicke. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:03:24 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:03:24 The meeting name has been set to 'vpp_weekly' 16:03:55 Hello all 16:04:06 tfherbert: hey there 16:04:23 joelhalpern: ! Hey there.... shall we discuss port-mirror abuse ? 16:04:55 DaveBarach: hey hey 16:05:07 Hey Keith... 16:05:18 #chair alagalah 16:05:18 Current chairs: alagalah edwarnicke 16:05:25 #chair DaveBarach 16:05:25 Current chairs: DaveBarach alagalah edwarnicke 16:05:30 #unchair alagalah 16:05:30 Current chairs: DaveBarach edwarnicke 16:05:32 #chair afredette 16:05:32 Current chairs: DaveBarach afredette edwarnicke 16:05:36 #chair alagalah 16:05:52 #topic Agenda Bashing 16:05:59 #link https://wiki.fd.io/view/VPP/Meeting#Agenda 16:06:37 #chair alagalah 16:06:37 Current chairs: DaveBarach afredette alagalah edwarnicke 16:06:40 morning everyone 16:06:53 #chair joelhalpern 16:06:53 Current chairs: DaveBarach afredette alagalah edwarnicke joelhalpern 16:07:24 Call for any missing agenda items folks would like to see covered 16:07:28 Please speak up :) 16:10:01 afredette: tfherbert not to put you on the spot or anything :-P 16:10:19 #topic Getting Involved 16:10:31 alagalah: Thats OK. 16:10:48 #info edwarnicke Sharing "Getting involved" page 16:11:06 #link https://wiki.fd.io/view/Main_Page#Get_Involved Getting Involved 16:11:58 #info edwarnicke implores folks to proliferate channels but please document them. 16:12:52 #info alagalah preference is keep convention of #fdio- be kept for official projects, and ##fdio- for anything else but... its just a convention 16:13:16 #info mailman hasn't been behaving and we have reverted from mailman3 to mailman2 16:13:56 #link https://lists.cloudfoundry.org/archives/ CF mailer lists 16:14:11 https://lists.cloudfoundry.org/archives/list/cf-dev@lists.cloudfoundry.org/ 16:14:33 #info edwarnicke notes that this looks like mailman3 which isn't working for fd.io 16:14:42 #info searchability is good. 16:15:11 #link https://wiki.fd.io/view/Project_Proposals Project Proposal 16:15:17 tfherbert: oooooh yeah 16:16:43 #topic Consumability 16:17:25 #info Upstreaming .. we have afredette tfherbert from RedHat who maybe able to give us insight on upstreaming ... and potentially if folks on from Canonical could make themselves known ... 16:17:35 tfherbert: Whats the process ? 16:18:06 Need to have a page with the proposed projects too and if they have been approved, just to make it simple to see them all. Did I miss this suggestion 16:18:15 KeithWiles: Great call 16:18:22 KeithWiles: Good point :) 16:19:41 #info tfherbert afredette Committed to making VPP work with CentOS et al but not 100% on the process 16:20:09 #action afredette (RedHat) to find out process/checklist for upstreaming RPM package 16:20:17 #undo 16:20:17 Removing item from minutes: 16:20:32 KeithWiles: We do have a section on the project proposals page for listing project proposals (and instructions to list your proposal there : ) 16:20:35 #action afredette (RedHat) to find out process/checklist for upstreaming RPM package. Next week. Email pointers to RTM good enough 16:23:38 BTW - the wiki says the TSC meetings are Thursday on the first logisitics line, and Tuesday on the second logistics line... 16:25:08 #info call for volunteers to getting rpms building as part of verify/merge 16:25:17 On my new install of 15.10 I had to revert to upstart from systemd 16:25:22 #info call for volunteers to get rpms pushed to yum repos are part of merge 16:25:44 KeithWiles: Any interest in making the deb packages smart enough to use systemd if installed on 15.10 ? 16:26:23 #info call for volunteers to get deb packages working well on Ubuntu 15.10 (upstart -> systemd, uio_igb -> uio-pci-generic 16:27:14 #chair ashyoung 16:27:14 Current chairs: DaveBarach afredette alagalah ashyoung edwarnicke joelhalpern 16:28:05 #info dmarion CLANG works 16:28:05 #info DaveBarach describes gcc 5.2 issues 16:29:13 #chair cj 16:29:13 Current chairs: DaveBarach afredette alagalah ashyoung cj edwarnicke joelhalpern 16:29:48 I used clang for DPDK all of the time, but it can have a minor performance reduction in some cases. I have not noticed the performance, but I have not been doing a lot of testing. 16:30:58 DaveBarach: Have we considered manpages too ? 16:31:07 DaveBarach: At least for CLI/API ? 16:31:45 alagalah: Good idea 16:33:19 alagalah: I think manpages is a part of a larger question 16:33:24 alagalah: We have two kinds of docs we need 16:33:28 a) Dev docs 16:33:30 b) user docs 16:35:03 In DPDK the .rst files do have a lot text to describe details beyond APIs in the doxygen 16:35:43 manpages are kind of a staple 16:36:05 And I use doxygen in every project I touch 16:36:24 Everything else is extra credit :) 16:36:27 edwarnicke: yes 16:36:46 Developers on, developers in, consumers (install;provision) 16:37:04 Just started working with .rst 16:37:48 #info we need docs for "vpp developers" "vpp api users (like control plane)" "vpp deployers (people who just want to install and run it)" 16:38:10 DPDK got away from wiki as docs because you need to be online to view, so all of the docs are in the project and viewable without a network connection 16:38:56 #info probably both a control plane and a data plane section at hackathon? 16:40:35 agreed 16:41:48 tfherbert: What sort of split do you think ? 16:42:27 tfherbert: Gut feel is fine... 30 CP/ 70 DP ? 16:42:34 20/80 ? 16:43:20 tfherbert: CP requires a fare amount of learning curve. 16:43:38 tfherbert: Depends 16:43:47 tfherbert: Yeah... I'd humbly suggest that VPP isn't the place to come learn CP, but if you already know it ... 16:44:00 tfherbert: I think about it as three pieces : deveopment in the dataplane, development in the managemenet agent, development in the controller 16:44:27 alagalah: I don't know what the split should be. They each have different issues. Right, but CP VPP specific. Dave covered it at session but we need more in depth I think. 16:46:06 tfherbert: Say more :) 16:46:19 Hyperscan is open source and could be used for a very good regex engine 16:46:49 #link https://gerrit.fd.io/r/#/c/327/ - tcp/udp checksum disabling patch mentioned 16:48:04 https://github.com/01org/hyperscan 16:50:25 joelhalpern: Why Joel, whatever are you referring to? :-P 16:53:34 #info joelhalpern points out we would be doing he world a service by talking about SFC SFF and SF correctly as we look at NSH 16:53:54 #info joelhalpern and edwarnicke spend some time agreeing emphatically about other subtleties ;) 16:54:02 Thanks Ed. 16:54:23 #link https://www.openhub.net/p/fdio -> commit velocity documentation 16:55:25 #info Very interested in the NETLINK project 16:56:16 +1 on Netlink project :) 16:56:25 mdgray: :) 16:56:56 mdgray: Is there a NETLINK proposal underway ? If not, lets start one... mdgray ? 16:58:16 #info: I agree on packaging first+1 16:58:24 alagalah: I could help review any project but I don't have much time to commit to it 16:59:25 #info https://wiki.fd.io/view/Mature_Release_Process <- link to mature release processes 16:59:32 edwarnicke : Is Honeycomb itself a project? 16:59:43 afredette: It will be :) 17:00:08 mdgray: You reviews are welcome :) 17:00:32 mdgray: Anyone can review patches :) 17:00:56 I would actually favor closer to a two month release cadence for VPP itself 17:01:01 Turn the crank fast and agile 17:01:57 alagalah: afredette: hackfest Control plane breakout would be largely about Honecomb? 17:02:17 tfherbert: dunno I guess.... 17:02:21 Is there any collateral available to reproduce the benchmarks that Maciek has done? 17:02:49 #info alagalah asks about how we number versions 17:03:20 #info alagalah suggests sematic versioning 17:04:41 #link https://datasift.github.io/gitflow/Versioning.html <- semantic versioning explained 17:04:51 edwarnicke: :) 17:04:56 #info other suggestions involved YY.MM versioning 17:05:53 #info edwarnicke No, I wasn't advocating it. I twas seeing Relase 1.0.0 on the Mature Relase Process and had questions 17:06:31 KeithWiles: Yeah... that was what I should have said to be clear... ABI 17:06:36 Within VPP 17:07:50 KeithWiles: Good with time based release numbering 17:08:09 Yes, I agree :-)\ 17:08:16 Thanks 17:08:18 Thanks everyone 17:08:26 #endmeeting