Operating System: Mac OS
Platform: PC

[08:43am] Madhu: hi michal_rehak
[08:44am] Madhu: michal_rehak: regarding https://git.opendaylight.org/gerrit/#/c/7607/
[08:44am] Madhu: we spoke to the other michal from of lib
[08:45am] Madhu: and the extensibility support is NOT available yet and we agreed to extend the NXM inline till it is made available
[08:45am] Madhu: we have been waiting for more than 8 weeks
[08:45am] Madhu: we need to get moving and would be great if we can get this merged in now.
[08:46am] Madhu: and once we have the extensibility support, we can of course remove these and move forward. does that work ?
[08:46am] michal_rehak: Madhu: hi
[08:46am] michal_rehak: Madhu: I spoke with Ed regarding this
[08:46am] Madhu: oh great.
[08:47am] Madhu: and...
[08:47am] michal_rehak: so
[08:48am] michal_rehak: having Nxm0 and Nxm1 classes in basic implementation of OFJava is ok, since those are metioned in OFSpec
[08:48am] Madhu: agreed
[08:48am] michal_rehak: what I have problem with are the underlying models like TcpFlag
[08:48am] Madhu: technically
[08:48am] Madhu: even for ofjava
[08:48am] Madhu: i didn't want it to be part of the library project
[08:49am] Madhu: because tcpflag is NOT defined in the spec
[08:49am] Madhu: though the class is vaguely specified
[08:49am] michal_rehak: but in order to have at least some things working it is a bad idea to wait for refactoring of OFPlugin and creating bundles etc..
[08:49am] michal_rehak: agreed
[08:49am] michal_rehak: so
[08:49am] Madhu: our request now is to get this moving just to make sure we hit the helium dates
[08:50am] Madhu: if we have the extensibility support in, we will certainly move it to a pluggable bundle
[08:50am] michal_rehak: for POC I agreed with this change
[08:50am] michal_rehak: right
[08:50am] Madhu: do u know when will we have the support for extensibility ?
[08:50am] michal_rehak: but once this has to be moved to separate bundle
[08:50am] michal_rehak: in OFPlugin?
[08:51am] Madhu: overall.
[08:51am] michal_rehak: good question
[08:51am] Madhu: the problem is
[08:51am] Madhu: we have 6-7 NXM extensions to be done
[08:51am] Madhu: and I hate to embed the code like this
[08:51am] michal_rehak: in OFJava it is ready for testing
[08:51am] michal_rehak: so there are 2 more ways to achieve this
[08:51am] Madhu: okay
[08:52am] michal_rehak: #1 - register translators for OFPlugin in order to use external code
[08:52am] michal_rehak: but there is large refactoring needed for that
[08:53am] Madhu: ok
[08:53am] michal_rehak: #2 - vendor specific messages can be forwarded through OFPlugin without touching them
[08:53am] Madhu: hmm then controller will be corrupted
[08:53am] Madhu: i mean the md-sal generic yang models
[08:53am] michal_rehak: why so?
[08:54am] michal_rehak: right
[08:54am] michal_rehak: any bundle can define yang containing augmentation for existing md-sal model
[08:54am] michal_rehak: so can the vendor bundle
[08:54am] Madhu: right
[08:55am] michal_rehak: this is current case - there are augmentations on sal models as placeholders for specail messages
[08:55am] michal_rehak: similar augmentation is done on OFJava-API models
[08:55am] michal_rehak: in order for OFJava to read those messages
[08:56am] michal_rehak: but OFJava does not need to read them - for this purpose there are registered codecs
[08:56am] Madhu: can u point to an example ?
[08:56am] michal_rehak: matchEntry
[08:57am] Madhu: i will take a look @ that later today
[08:57am] michal_rehak: in md-sal there is a list of containers
[08:57am] Madhu: just curious
[08:57am] Madhu: if we can send extensions transparently to ofjava
[08:57am] Madhu: why do we need any kind of translators at all ?
[08:57am] michal_rehak: well the main point is that you can not have one augmentation able to augment 2 different nodes
[08:57am] michal_rehak: but augmentations can share grouping inside
[08:58am] michal_rehak: that is my point!
[08:58am] michal_rehak: #2 will completely skip messing with extensible messages in OFPlugin
[08:58am] Madhu: I need to refresh my yang knowledge
[08:58am] Madhu: #2 is cool. but am trying to understand ... at what cost ?
[08:58am] michal_rehak: so for POC this is huge delay
[08:59am] Madhu: i agree that the current change is PoC due to the lack of end-to-end extensibility support
[08:59am] Madhu: and we have 7 more coming
[08:59am] Madhu: so not sure how to address this sooner.
[08:59am] michal_rehak: we need new bundle defining those augmentations in order to interconnect md-sal and ofjava-api
[09:00am] michal_rehak: and vendor bundle - this we need anyway
[09:00am] michal_rehak: and minor changes in OFPlugin in places where those extensions will pass
[09:00am] michal_rehak: OFPlugin will not be dependent on vendor bundle
[09:01am] michal_rehak: vendor bundle will be accessible for NB-app (or those could be the same bundle)
[09:01am] Madhu: okay.
[09:01am] michal_rehak: and vendor bundle needs to register codes for OFJAva before sending or receiving extension messages
[09:02am] Madhu: am okay with any idea if we have a solid end-to-end idea to work from my app towards controller -> ofplugin -> ofjava
[09:02am] Madhu: and back
[09:02am] michal_rehak: and this way md-sal models will also stay vendor agnostic
[09:02am] Madhu: and would be helpful if we have an example done by you experts so that we can follow that
[09:02am] michal_rehak: right so
[09:02am] michal_rehak: here is the proposal:
[09:03am] michal_rehak: the first steps / POC can be done just like Brent already pushed but with HUGE comments that this is about to get changed
[09:04am] Madhu: how about a bug opened with these gerrits to be backed out ?
[09:04am] michal_rehak: and at least all augmentation definitions shall be pulled in separate yang files in order to easy the later changes
[09:04am] Madhu: +1 to that
[09:05am] Madhu: so. Brent is working on the current code to get the OpenStack integration going. and hence I wouldn't want to block him
[09:05am] Madhu: but I can take this cleanup work
[09:05am] michal_rehak: ok
[09:05am] Madhu: but I need help from you guys to make sure we have the infra in place for it to work
[09:05am] michal_rehak: and is Brent around?
[09:05am] michal_rehak: sure
[09:05am] Madhu: unfortunately. he works nights and nights
[09:06am] Madhu: and goes offline in the mornings. lol
[09:06am] michal_rehak: first I need to write a wiki for that
[09:06am] michal_rehak: oh
[09:06am] Madhu: yes. thats a plan then
[09:06am] michal_rehak: ok, I will reach him tomorrow
[09:06am] Madhu: perfect.
[09:06am] michal_rehak: and then we can file bug a check planning
[09:06am] Madhu: if we can get Brent moving with this PoC implementation, that would be awesome for the short term
[09:07am] Madhu: will file a bug right away infact
[09:07am] michal_rehak: currently I am occupied by high priority task - ingress backpressure
[09:07am] michal_rehak:
[09:07am] Madhu: np. if u can unblock brent by +2 and merge that code
[09:07am] Madhu: we will actually start to integrate Security group Openstack feature
[09:08am] Madhu: michal_rehak: also, am new to md-sal yang. so please go easy on me
[09:08am] michal_rehak: everyone is new to yang as it is used in md-sal
[09:08am] michal_rehak:
[09:08am] Madhu: ok perfect so am not alone to be stupid
[09:09am] michal_rehak: ok, I will add some info in gerrit about the commenting stuff
[09:09am] michal_rehak: YEAH - we are never alone!
[09:09am] michal_rehak: and I will copy this conversation into mail fro brent
[09:09am] Madhu: ok sir. thanks a ton.
[09:10am] Madhu: excellent. sure.
[09:10am] michal_rehak: you are welcome
[09:12am] Madhu: michal_rehak: btw. since Brent is offline. I can amend the gerrit with the HUGE comments as we discussed
[09:12am] Madhu: can u merge it with that done ?
[09:12am] michal_rehak: yes - doing that
[09:22am] Madhu: michal_rehak: https://bugs.opendaylight.org/show_bug.cgi?id=1133
[09:22am] Madhu: opened a bug against our project
[09:23am] Madhu: and we can track all the gerrits that will introduce these PoC changes for the NXM extensions (we have 6-7 of them)
[09:23am] michal_rehak: Madhu: great
[09:25am] Madhu: michal_rehak: please let me know if you want me to amend the commit with some comments pointing to this bug ?
[09:25am] Madhu: you can also add an attachment to this bug with the conversation that we just add to keep me honest
[09:25am] michal_rehak: oh
[09:26am] abhijitkumbhare_: sorry my MAC hung up and had to reboot
[09:26am] abhijitkumbhare left the chat room. (Ping timeout: 276 seconds)
[09:26am] abhijitkumbhare_ is now known as abhijitkumbhare.
[09:26am] michal_rehak: the other Michal from OFJava knows about this issue
[09:26am] michal_rehak: Madhu: may I ask you to append the conversation to bug?
[09:26am] Madhu: michal_rehak: sure i will
[09:26am] michal_rehak: Madhu: thank you

Adding additional conversation to bug regarding models etc. Will need to figure out how to do this w/the Yang headers. Everything I see upstream is contained in single files, match, ext, action etc. Thanks.

[13:47:07] <michal_rehak> networkstatic: so if possible, please put vendor extension models into separate yang files so it will be easy to pull them into new bundle
[13:48:01] <networkstatic> in ofJava Im guessing?
[13:48:09] <networkstatic> Controller using other objects
[13:48:13] <networkstatic> existing
[13:48:13] <michal_rehak> networkstatic: right
[13:48:16] <networkstatic> rogr
[13:48:24] <networkstatic> you got it. will break them out
[13:48:36] amit_ (~amit_@2001:420:305:1252:b594:9ace:d657:9616) joined the channel.
[13:48:55] <michal_rehak> networkstatic: in controller there is for example by match simply list of fields (containers)
[13:49:37] <networkstatic> i want to use as little as possible from other projects to be honest
[13:49:39] <michal_rehak> vendor stuff among those should be one day moved out as augmentations... but this needs some designing
[13:49:44] <networkstatic> so you get us support
[13:49:50] <networkstatic> to use exteernally
[13:49:57] <michal_rehak> networkstatic: appreciate that
[13:50:02] <networkstatic> and we wont need to be in ur project
[13:50:13] <networkstatic> its no fun for us
[13:50:19] <michal_rehak> funny is, that nxm0 and nxm1 are part of OFSpec
[13:50:27] <networkstatic> last thing i want is to intrude
[13:50:39] <networkstatic> well, i agree
[13:50:49] <networkstatic> why i think its reasoanble
[13:50:58] <networkstatic> since u defined it in the clazz