ONNX Foundation WG, 10 Sep 2019

Previous Meeting

⏯

youtube image

►

From YouTube: ONNX Foundation WG - Eclipse intro meeting 20190910

Description

Meeting recorded by Jim Spohrer (IBM) on September 10, 2019. This is intro meeting with Eclipse Foundation: Mike Milinkovich, Paul Buck and Daniel Bandera. Ryan Loney (Intel) also joined.

A

A

Hi this is James borer. This is our onyx working group called for onyx and a foundation. Today's call is with the Eclipse Foundation and thank you Mike and Paul for making time Ryan is my co-chair of the working group that the onyx onyx team community has started to investigate different foundation home possibilities for on it.

A

So, thanks for making time, I did send out a little agenda that we would like to go through, and the start of the agenda is just brief introductions so I'm again I'm Jim spoor from IBM and one of the co-chairs of the of the Kea working group and Bryan. Did you want to introduce yourself, Nick, yeah sure, I'm, Ryan money, I'm product manager at Intel, working on our AI, imprint accelerator software and excited to to hear more about the different foundation options and I'm glad you guys took time to come present to us today?

A

Thank you and Mike and Paul. Do you want to introduce yourself go ahead? Well, yeah.

B

So, pull about VP of community development at the Eclipse Foundation focused on new communities, growing existing communities and overall healthy organization and.

C

I'm Mike linkage the executive director of the equals salvation perfect.

A

Thank you like them done great thanks, yeah, we I think Mike. You and I have had a few conversations about AI in the equipt foundation in the past. So thanks for so quickly agreeing to have this call when I reached out about the Onyx community through a very exciting community that said, Microsoft, Facebook and Amazon started to create a representation for working and machine learning, graphs and represen, but not looks like add-ons right. Well, yes, I am really no, no worries we're just getting introduced. Adam.

A

Do you want to introduce yourself and by the way, we're recording the meeting? Because there were several people from the onyx community who can join yeah.

D

Chorus fun, so I'm adam the co-founder skyline representing the deporting for today, project free clips and kind of helping with some of the AI stuff happening of the foundation. Okay, fantastic.

A

Thing, and so mike again, I think you know a lot of people in the onyx community. Certainly the people from IBM and and some other places, no, the eclipse foundation. But we thought we would. You know just go through some basic questions for community and if you could just share a little bit of the history of the Eclipse Foundation and you know what what its mission is for. No.

B

I'm, not boarding.

A

New types of projects that would be really helpful, start yeah.

C

Sure happy to do so so because Foundation has been around for 15 years, I think we can actually I think we might be the very first 501 C 6, open source organization. That's still around, and originally our roots were in software development tools to this day to a large number of typically Java developers. If you say the word Eclipse, they think oh yeah, the Java IDE.

C

However, as you know, other institutions have done. We have evolved over time to become a very large tent hosting a great number of open source projects and technologies up to about three hundred and seventy-five projects.

C

Today, 275 organizational members of the Eclipse Foundation and our technologies span from our origins and tools to advanced simulations for testing artificial intelligence in autonomous driving to deep learning for Jay, which we mentioned, of course, a pretty strong presence in cloud native Java through our initiatives in Jakarta EE and the micro profile, vertex jetty and the likes, and very strong in IOT as well.

C

Arguably, the center of gravity for open-source and IOT and edge with almost 40 different projects and about 3 million lines of code focused on on IOT and edge, so a very wide, very big tent of projects. Starting about a decade ago, we started organizing what we call working groups which provide focus on particular topic areas. Each working group is a self-governing consortium order under the banner of the Eclipse Foundation, with its own charter, its own steering committee. The the projects have to you know, follow the Eclipse development process.

C

However, there's a wide this large degree many degrees of freedom in terms of how the each working group can govern its its affairs.

C

Starting with about two years ago, when Oracle announced that Java EE was going to be moving to the Eclipse Foundation, we became involved in creating a new and modern specification process based on open-source principles for use by Eclipse projects and working groups and I think that's actually a feature that so particularly interest for by the way I'm sorry use it. You guys pronounce it onyx or Owen and X, or how do you actually say it? I I always say onyx, I, hope.

A

I'm doing it right yeah, our lawyer stated.

C

A

But we say onyx yeah.

C

I'm not surprised yeah, okay, so when so I you know as I understand it. Onyx is primarily a specification, that's its major major raison d'être and we can help support that through a specialization of the consolidation specification process. So the process that we came up with was by design a template which is intended to be refined by each specification team that wants to adopt it, and so it's I think it would provide an excellent template for for onyx. It has all the attributes that you would expect from an open specification process.

C

We've refined all of our licensing agreements, including all of our contributor and committer agreements and our IP policy, to make sure that patent rights flow appropriately to independent implementations of the specification we've crafted a very liberal, equal edition specification license that's modeled after the w3c license and for assuming that onyx has some sort of testing capabilities for compliance to some sort of TCK org like we also have a TCK binary TCK license to make sure that when people are making compatibility, statements and branding statements that they're basing it off of the correct version of the TCK, the assumption in our spec process is that all of the artifacts under development are maintained in open source under either the EPL, the Apache or the BSD license.

C

And but then, once you actually go final with a specification and a TCK, there are good reasons to use licenses other than open source licenses for those final versions of those documents and artifacts. So and we've got experience, you know that. Actually, the timing is excellent, because this morning Jakarta ee8 went live. We're done. We can now say that we have shipped a very, very large and important set of specifications to the industry through this process. So we think that we've demonstrated its value in utility.

A

Very great, so the main reason why onyx is searching for a non-profit foundation. This point is various contributors and members of the community or starting some equipment plans. Product roadmap plans that include on it and so rapidly getting the IP and assets into a vendor-neutral home can own the IDN assets is, is the main reason so could you say a little bit about the ability of the Eclipse Foundation and the process? How long does the process take?

A

You know for organizations like Microsoft and Facebook and others to contribute code and IP and assets to to do a work working within the Eclipse, Foundation or project in the Eclipse Foundation. Well,.

C

Based on 15 years of experience in this job, I can tell you that it's exactly as fast as your lawyers want it to be.

C

Know it more seriously, you know it is perfectly realistic to think that you could establish a working group with a specification process. That's completely functional, you know within a couple of months it really, but it really depends on the details and it particular depends on how can I put this the goodwill of the parties and the lawyers involved? You know if there is one you know fully engaged party that, for whatever reason is in their interest to cause this delay.

C

You know it. Every open process is a. It can run. Afoul of I, wouldn't even call them a bad actor, let's just say a deeply self-interested actor, but the process is specifically designed to be to be fairly quick right and fairly lightweight that that is, that is, that is the design point of the process and so the for example. If you look at the Jakarta EEE specification process, it is a specialization of the Eclipse Foundation specification process.

C

Template I think it's something less than twenty lines of text right, so the specializations can be very clear and very quick. Yeah and all of the contributor and committer agreements are set up. The membership agreements are set up. We have a participation agreement, template it could be fixed up then literally a day. So there's the project is similarly for a draft charter for the working group. You know we have all these artifacts. The process can absolutely run very quickly if everybody is working it together to make that happen. Yeah.

A

I think that's the case. It seems to be the case. I think it's a very well-run community, my perspective and everybody can be very well aligned but I'm curious, Microsoft, Facebook, Amazon, Intel Nvidia. Those are the five on the steering committee right now are all their lawyers quote-unquote trained up on the Eclipse Foundation? Is there? Have you done things with all of those five firm.

C

So not all five, so IBM and Intel I would say, are quite well trained on all of our processes. Intel, as we speak, are, is in the process of joining two additional working groups. So you know their attorneys are going to be quite familiar with our participation agreements and so on. So I think that that those two in particular are well are well trained.

C

Microsoft is an existing member and has been a member of the Jakarta EEE working group for quite a while and has participated in many of the activities there, so certainly within some portions of their team they're familiar with with what we've got. Unfortunately, neither Facebook nor and VD are members of eco foundation. As of this moment about.

A

A bizarre did you say: Mike did you say Emily, oh.

C

Sorry uh answering Amazon, Nvidia and Facebook okay, none of those three are members of, although like I can't get into details, but let's say that one of those three is actively reading. All of those the same materials for different reasons, understood.

A

I think it's interesting that you know you've had a long relationship with Intel and IBM and Ryan.

B

A

I are the co-chairs of this group, so that says something I guess so I guess the the next question we've been well trained by our lawyers to be efficient, getting into nonprofit foundations. I think is a punch line here. So onyx community, as I said, is self-governing. It's made good towards open governance. There's good good management cuts during committee. Good communication, good, you know everything I I have to say the community is self-governing in a very, very positive way.

A

I think the biggest complaint is, you know, pull requests aren't getting merged fast enough and being review. That was, we felt the only complaint. I heard, don't ask you could be needed, which is a standard complaint. So how does you know is the ability for onyx community to stay self-governing between comment a little bit about the lydian process? For that happen, my so I'm.

C

Assuming that onyx today is basically run as a github repository for.

A

A spec right, that's my cry! Not.

C

A

Right, that's correct, yep.

C

Do you have a so good yeah.

A

There's a lot more involvement than yes, so.

B

We do have regular.

C

A

And face-to-face contact and online yeah sure.

C

But is there a CLA in place, yes,.

A

There's a ceiling I guess we needed.

B

A

Have a foundation for a legal entity.

D

A

There's one key yeah: that's one of the things we would like to move away from is the CLA I believe. There's several members who said they'd like to move away from it once in a non-profit foundation. Is that did I say that right, Ryan, yeah, yes, I was needed, but yes, that's that's. That is the preference of our lawyers. But if you know, if we aren't able to move away from a CLA we we would like to have a licensee that is an onyx foundation, a non-profit, so yeah.

C

Okay, so the wait so they and what is the license, the contribution and distribution license at the moment. Is it Apache.

A

Yeah I believe so you know Ryan for sure. Okay, we got two passes as well: Mike.

B

I think that I don't.

A

Think that it is and I think that was one of the issues from from legal is that it's not currently.

B

A

Notice, let me go check right, you know, okay, so.

C

The way the Eclipse Foundation is set up is, first of all, about half of our projects and activities are hosted on github. So we're very familiar with that. We do. We do insist on contributors signing the Eclipse contribution agreement, which is basically a wrapper around the Linux foundations. Dco and, basically, all it says, is so there's no rights given to the Eclipse Foundation per se by contributors.

C

It's basically it's just an agreement that says that we know we're making the contributions under the project license and that when we sign off our contributions that we you know, we are signing off on the DCO and and here's our information for your record keeping that's all as the licensee would be the Linux Foundation.

C

No, no, the licensee is the lights that the licensee is recipients of whatever contribution is so there's no assignment of any rights to the Eclipse Foundation. We assert ownership over the completed specification document as part of the our spec license, but we're basically saying that we own the over the collective copy work on the on the spec document. Since we hosted all of the processes that that were involved in putting together the specification license.

C

Yeah so I noticed I'm, Emily, key lights, the lights out.

A

I might yeah okay, yeah.

C

So, okay, so from my purse, so there are many lawyers with many opinions about how these kinds of things need to be put together.

C

So the the advantage of the apache software license over the MIT license is that the patent license is royal is is explicit and royalty-free, however, and the apache license is not bad as a spec license, but there's basically no there's no way that if you have specification published under any open-source license the the license grant specifically say that you can change the specification itself right. That is actually what you know. That's an attribute of any and all open-source licenses.

C

So when our our spec license says is so in the only time we use it is when the spec goes final. So you know all of the attribute. You know everything works under open source until you actually hit the final version of the skirt, a version of the spec that you want to release as fine up and then, when you do that, then there's we put on the spec license, and it says this thing is copyright, Eclipse, Foundation and anybody is free to implement this using whatever license they want, including proprietary.

C

But you cannot change the spec document itself and we actually think that for specifications, that's actually a pretty good way to go. The other thing I'll say is that you know again, maybe the lawyers that were participating in in our discussions and in their lawyers from you know: IBM, Red, Hat, Oracle and and others.

C

Maybe they were particularly conservative, but in the construction of a specification whose goal is to enable independent implementations, I've always heard it argued that it's to everyone's advantage that there be a some form of CLA, some form of accurate record-keeping on who are the Ordway of where are the origins of all contributions, etc, etc. And so that's why we've constructed things. The way that we've constructed them.

C

I would also mention that I sorry I forgot when I was talking about the equips contributor agreement beforehand, one of the things that we added when we were constructing our spec process. That I forgot to mention is that our contributor agreements now say you know anything any copyrightable materials that you contribute to a specification.

C

You know you permit the Eclipse Foundation to relicense that under the Eclipse Foundation spec license. So that's part of our standard contribution agreement. Again we have we've spent a lot of time, basically optimizing open source workflows that still allows you to deliver. What even the most conservative of lawyers would consider a very well protected, well license and well license specification document which will enable independent implementations.

A

Okay, great I, think those are the major questions. There were additional Q&A that I had like: where is it already Onix adoption or future potential energy foundation project? And maybe you know oh um yeah,.

C

Maybe if, if Adam can jump in here, Adam is the project leader for the Eclipse deep-learning 4j project, so he's going to be more familiar with onyx than polarized so Adam, please take it away. Yeah.

D

So, um actually, as a community, we're actually heavily involved in TF Java Karis, as well as implementing onyx support as well, so we're kind of like, but we're kind of more aiming for thinking with the niche Cafe two names for a new spark, Apache spark and kind of big data ecosystem.

D

So I think you know there's a lot of you know: I'm, officially going to sell a little bit of kind of Eclipse Foundation and what they've done for us in terms of you know, AI projects and a lot of the freedom that they've afforded us as a community they're, allowing us to start looking at different things like working groups and different training, different synergies across actually the application stack in the enterprise, so I think um in terms of Onyx itself. You know we'll be adding support for onyx as well.

D

We mainly have concert oil import right now, but we're in Paris import but we'll be adding will be adding onyx import as well. So that's one area and I think bridging into like real enterprise. Adoption I think eclipse could really help streamline that. So you know, they've they've also really bridged. Well and they thinking to be kind of the cognitive eco system, you know kubernetes a lot of that.

D

A lot of that overlap, so I think if you want mainstream adoption of Enterprise, while still working with a foundation, that's able to deal with large companies, I think moving into the mainstream AI adoption I think he clips we keep than that. So the application stack is where clips I think is very strong and why we pick the Eclipse Foundation for deal for today itself.

D

We, we actually weighed the Linux Foundation as well, because ooh catchy foundation and we went with Eclipse for the the reason that you know not only do they have been the enterprise adoption, but they're they're also a lot more familiar with the the software stack for developers and because that's kind of our niche. That being said, like I, said they bridge will as well into other parts of you know. Other parts of the ecosystem wouldn't know where from so enterprise developers was definitely the main audience that we were targeting here.

A

Got it right and I know Dan bandera IBM mean he told me to look at a I work with the auto industry. I think that's going on an eclipse foundation talking to be synergies there I, don't know if anybody components that or not yeah.

C

So in particular, there are two working groups that have started up in the last couple of months: open Genesis and open ATX, so open open Genesis is the prime mover behind that is an organization called tube sued, which is a major part of the certification and safety qualification infrastructure in central Central Europe, and the mission for open Genesis is to create an infrastructure, an open source infrastructure for safety, qualifying autonomous driving systems right.

C

So you know, which is obviously ultimately about safety safety qualifying, AIS and and neural networks. Now and the other one is open.

C

Edx the prime movers behind that are Bosch with IBM Red, Hat, Microsoft and others, and what its wants to build out is a end-to-end tool chain for building autonomous driving systems, which things like safety, qualification data protection, data, replication and so on are all going to be an integral part of the tool chain and when I say tool chain, I don't just mean like an IDE I mean everything from verification and test analysis system requirements gathering you know the whole the whole enchilada for delivering safety, qualified systems of high degrees of complexity, yeah.

A

Great. um Thank you. The one last thing: it's there that there is some onyx. Let's have some core values. We want to look them up right now, Ryan I, don't know if you have it handy, but I know that they're, basically, principles of speed and openness and transparency and all of the things I imagined. That's in order, bread and butter for for the adults foundation down.

C

So that sounds like standard operating procedure around here. Yes,.

A

I, certainly like Bend and are before I meet. My IDM colleague has completely booked foundations. That means talking a lot about that. So I think that's really all we have I just so. You know what our next steps are.

A

We we're doing this interview with units from the major foundations that are being considered, which are our clips apache linux foundation, ai and we're all also considering ml per ml commons for those before major ones that it looks congested so far, I'm like I was just informal conversation the beach, so that we could gather some of the basic information and take that back to the steering committee and then there's going to be able to be some boating and things like that.

A

But I imagine you know we want to reach out again and we do plan to post this report written on. We are working the onyx working with website or anybody in the communities you can go in complete it. So, thank you so much for letting us have this opportunity, good activity today and record this and and buying it, and if you have anything else, no I think that's good and yeah we're going to share the recording so yeah thanks Mike intakes Adam, really appreciate you guys taking the time yeah door.

D

A

So expressly we'll be circling back, but thank you very much for this initial information really super helpful. Thank you before the follow. Thank you. Oh thank youumm. Walker now.