After lunch, it was time to come back in the “Open Source for industrial users” track lead by Gaël Blondelle.
Increasing industries speed to innovate with FLOSS by Dominique Toupin, Ericsson
- Dominique started by asking a question: Does speed really matter ?
- He rapidly concluded that yes, of course. He gave some examples of projects initiated by Elon Musk, such as Zip2 sold to Compaq in 1999, Paypal. Or Tesla (Electricity car) and also SpaceX. All were very complex systems elaborated in a short time thanks to Open Source. Same is true for Google/Android.
- You end up with better features by doing Open Innovation and teaming up experts from different companies.
- This is also valid inside your company: whole greater than the sum of the parts. And you’re not locked in.
- Only 15% of RFE are really implemented in commercial products. In FLOSS, when a feature is key, you can do it yourself or buy someoneelse’s time so that it is realized at 100%.
- People tend to oppose FLOSS to commercial, make to buy. It’s not the case. FLOSS is commercially supported, so just take the best of both worlds to fullfill your need of speed.
- Requiring tools across the whole chain (and expensive ones) slow down your service activity, whereas using FLOSS tools in development brings speed to the service part. And you gain time with existing knowledge from universities or company acquired.
- FLOSS allows to dedicate the extra budget gained on licenses costs into the features you need.
- E/// has an Open Source Core team.
A very pragmatic approch exposed by Dominique, showing clearly tradeoffs needed at industrial level.
Efficient and safe FLOSS strategy by Michel Ruffin, ALU (on behalf of Philippe Richard, VP of Corporate CTO)
- Size matters: 79000 employees, 27900 patents, 27000 developers, 130 countries, numerous suppliers and outsourcing, multiple acquisitions per year (=> deal with legacy), life cycle from 1 to 20 years. Makes developing the Governance process “interesting”.
- Trend towards becoming an integrator of FLOSS with more complex SW stacks, reducing however the development costs during time.
- ALU’s strategy is going to FLOSS to remove supplier lock-in, much more than to reduce costs.
- Between 20%-80% of FLOSS components in their products (40% in average). Importance to create internal communities to discuss FLOSS related topics. FLOSS adoption means innovation, speed, freedom, new business model (moving from a HW/SW supplier into a service supplier)
- ALU is a contributor of FLOSS (even if not known). By paying providers (10+MUSD), providing patches/bug fixes to tools, Corba/Mico, Plan9. Also sponsoring OWF, FOSSBazaar, Systematic, OVA, Carrier Grade Linux (LF).
- For ALU, it matters to respect the philosophy behind the words of the license and thus contribute.
- Strong FLOSS Governance process started in 2002. Process evolving constantly (taking in account new techno/licenses/acquisition/…) 160 people trained 1 week to be FLOSS validators. 1000 people trained on a basic tutorial. 3500 FLOSS components in ALU DB. Clauses in supplier contracts (propagation to their own suppliers). ALU willing to share the governance process with other companies. ALU would like to standardize these clauses with the Compliance group of the LF.
- R&D is declaring FLOSS usage. ALU is also automating the BoM by scanning code (BlackDuck/protex and FOSSology)
- All this is available as much as possible on the Internet (However, I was not able to find easily the oprtal mentioned in Michel’s slides :-()
- On top of the process, you need to check that it’s applied (start with CxO, R&D – even if they think they know), Communication). Then improve the process, deal with exceptions, stay flexible, and stronger during time.
- Resources to support the process needs to be allocated accordingly. Use tools to automate and to detect issues and inform executives.
- Challenges around stuff like Maven, SPDX adoption, partnership with other companies …
ALU presented a strong Governance model, including now suppliers, and is willing to share best practices with others in order to improve the ecosystem. Network Equipment Providers are clearly taking seriously this area.
Business model of co-development on FLOSS by Denis Pillat, Service Delivery Manager for ALM at ST Microelectronics and Laurent Charles, Enalean
- Custopmer (ST) funded the development and save on the maintenance by contributing to the product Tuleap (a FLOSS ALM).
- Customers’ developments are also supported by the partner (Enalean).
- Strong internal usage of the forge (120000/40000 users) so central, with requirements around robustness and availability (ran 24×7) and long life cycle, but with an improved TCO. If budget is cut, needs independance from provider.
- ST is not an ISV, team role is to support deployment and integration in ST landscape.
- Solution retained is a mix of in-house and outsourced solution.
- Using and adapting a FLOSS costs as it requires backporting features each time with new versions, and ST is not scaled to cope with the rythm of a FLOSS project.
- Code and features from ST are reviewed with Enalean so easy to integrate. The partnership is of good quality. And also good quality of contributions.
- For ST, FLOSS increases motivation of contributors with their work recognized and exposition, and they work more on creative parts, and less on maintenance tasks.
I think the presentation would have been more effective if ST would have been the only speaker (or speak more). The track isn’t aimed at promoting companies, but really share return of experience around FLOSS adoption.
- Pierre first presented some characteristics of an airplane development in size throughout the years:
- 4 kB for Concorde
- 4 M for the A320
- 12M for the A330
- 500MB for A380
- Life cycle: 40 years – A300 family (started in 1972 and production stoped in 2007 and support till 2050 = 78 years). Tools need to be there for a very long time.
- Code is increasing. Quality is mandatory
- Historically, development of their own tools to check quality. Not their core business. Moved to a buy approch.
- Internal tools transfered to editors, who tried to sell them on larger scale, which failed as too costly and too specific. Some examples:
- For Autan (Airbus name) => Attol (Marben) => Attol (Attol-Testware) => RTRT (Rational) => RTRT (IBM)
- For RTRT they succeeded, but Airbus has anyway problems with the life cycle of this tool.
- Scade (Airbus + Schneider) => Verilog => CS => Telelogic => Esterel Tech.
- Geode (Airbus) => Verilog => Telelogic even died !
- no more control on these tools by Airbus anymore. Sometimes can’t even buy a license anymore.
- Topcased started in 2004. Reduce dev costs using model based System Engineering.
- Integrated universities and academic partners.
- Topcased aims to produce tools for embedded domain on critical system, on the descending branch of the V life cycle.
- Community around topcased includes Airbus, CS, CNES, Thalès, EADS, Atos, AdaCore, INSA, EnSEEIHT, Toulouse Univs, Inria, Irisa, Laas, Onera at start. Now additional new partners such as Turbomeca, Continental, Obeo, Carnegie Mellon, CEA
- 2006: First FLOSS release. (One year to solve licensing aspects)
- 2007: V1.0 and then one major version per year synchro with Eclipse. Minor every 2 months.
- 45 subprojects from model editors to code plan generator, model simulator to property generator.
- 2011 first TopCased conference (> 100 persons)
- Allows competitors to work jointly on components.
- 12 components are in use today (A350)
- However, an organization is missing to improve quality and IP control, maturity assessment, VLTS build system, roadmaps. OPEES (ITEA project) aims at fixing that.
I really like this presentation (that I first heard partly during the Think Tank 2010). It clearly shows the huge problems that software development still needs to solve in order to support such life cycles. Raises questions such as how to motivate a community to maintain software for so long time, typically. Also how to preserve build environement, especially when the hardware is changing as rapidly as it is today.
It was then time for me to change session and move to the Governance track lead by Martin Michlmayr.
I contributed briefly to a join talk with Antelink.
Tools for developers to ensure legal integrity of their code by Freddy Munoz, Antelink and Bruno Cornec, HP.
Freddy explained in more details what Guillaume covered in his talk of the morning, and went through the details of Antelink Notifier, Reporter and Search. For myself I covered rapidly FOSSology, giving its main features and also the latests developments realized. Of course, as the project is hosted by the Linux Foundation, as long as they keep the systems away from Internet for forensic, it will be difficult to have access to the project :-( But hopefully, it will be back soon.
Identify the obligations of FLOSS by Benjamin Jean
- a License (or contract) is a tool made of rights and obligations, a scope and trigger
- Writers can be foundations or Companies
- Number of licenses increases (70 referencesd by OSI, >50 by FSF, 1000 by Black Duck, 400 FOSSology)
- Benjamin gave some statistics:
- For Black Duck 43% is GNU GPLv2, 11% is MIT, 7% is Artistic
- For OpenLogic 32% is Apache, 21% is LGPLv2.1, 14.4% is GPLV2
- We need clarification: a common nomenclature (detailed and scalable) & descriptive
- International standardization body is a good way, but very expensive, and not driven
- Benjamin proposes a first classification based on obligations (to give, to do, to not do.)
- Rights are harmonized across definitions (some more rights depending on licenses or some missing)
- The real differences are around trigger, scope and obligations. Benjamin then detailed those:
- Obligations have no common definition whereas a standard would be useful for projects, industry
- Scope can be very limited (permissive), limited (GPL/GPL sometimes, CeCILL-C, MPL), standard/legal (EPL, EUPL, OSL) or large (GPL, CeCILL)
- Trigger: Distribution (GPL), Usage (RPL), External deployments (AGPL, EUPL, …)
- License compatibility could also be classified between limited and extended. Cf also work done at the Inria, described on their Web site in the Innovation part, Free Software then the guide.
- This classification can also easily be valid across countries and thus not being dependant of local legal rules.
Benjamin’s approach was extremely sharp and that session was really deserving belonging to the ‘Think’ part of the OWF ! This approach by obligations could really improve the situation of licenses compatibilities and help all the actors of our FLOSS ecosystem.
I had still a bit of time to discuss with him, Martin and Marc Picornell before leaving the event and benefot from the fact I was in Paris to attend a concert at the Chatelet Theater performed by the National Orchestra lead by D. Gatti. Ravel, Dukas, Debussy and Enesco made a radical change for the end of the day !
You can see some of the pictures took during OWF 2011 at https://picasaweb.google.com/112434061686721373729/OWF2011