This is a report of the ODF Plugfest that was held in Paris on November 16 and 17, 2016. All the presentations can be downloaded from:

http://odfplugfest.org/2016-paris

This is the web version of the report.

The twelfth ODF plugfest was hosted by the French Direction Interministérielle du Numérique et du Système d'Information et de Communication d'État (DINSIC) , with support from Logius and organised by OpenDoc Society.

12th ODF Plugfest in Paris

Day I: Wednesday November 16th 2017

Location: French Ministry of Finance

The 12th ODF Plugfest was opened with a welcome by Laure Patas d'Illiers from the CIO department of the French Ministry of Finance, on behalf of the local hosts: the French Ministry of Finance and the Paris Open Source Summit.

Michiel Leenaars, vice-president of OpenDoc Society, gave an introduction to the plugfest. Delegates came from Australia, Canada, France, Germany, Italy, Russia, UK, Ukraine, Taiwan, The Netherlands, Luxemburg, United States and Belgium representing application vendors, standards bodies, public sector and industry. This included several governments as well as the European Commission and the European Parliament.

After a short introduction round, dr. Ben Martin (OpenDoc Society, Australia) gave a presention about his test results from work done since the previous ODF Plugfest in The Hague. He highlighted that significant effort lies in discovering patterns behind breakage of certain applications.

Jos van den Oever (Logius/Ministry of the Interior, The Netherlands and co-chair of the ODF TC) described the transformation made during the history of the ODF Plugfests, moving from manual testing at the first plugfest to structured automated testing. Subsequently he gave an introduction to the new testing infrastructure that has been written to support the creation of ODF autotests. It allow for elaborate tests to be created in an convenient web environment, with a lot of functionality to determine where certain problems are located. Each test consists of elements inserted in a document template, and a set of conditions that are checked after the document has been opened and saved by each application. In addition to automatically running all the tests through the various applications, the server also validates the resulting code to check if it is conformant with the ODF specification.

Prior to the plugfest, a significant amount of tests was created. During the event itself the focus was on triaging these tests, to see if they test the right things in the right was, as well as creating any new tests identified as missing.

Attendees asked how to best make the scores of applications in the different tests be fairly represented. The answer was that there is no such thing as a simple score, there are many different tests and people can themselves develop tests. Test are automatically validated before they are allowed in, and can be triaged by every registered participant. The scoring mechanism is simply a summation of the results in different sets of test documents.

Participants requested if mobile version and web versions of Google Docs could separately be added to the tools. The answer was that indeed this was a good idea. Jos van den Oever showed how to add new applications, and invited people to contribute to this effort.

The question was raised whether we should we develop a test similar to the "ACID tests" for ODF. The ACID tests were a series of tests that used the visual characteristics of the CSS standard to create a graphical representation of support for that standard.  For the short term, the focus is on creating many different modular tests as the raw primitives for more elaborate tests. In the longer term, the success of a visual representation is hard to argue against.  Indeed, some work in this area has already been done for ODF, and work in this area would be very much encouraged.

Franklin Weng (Software Liberty Association of Taiwan, Taiwan) raised some interesting issues with CJKV (Chinese, Japanese, Korean, Vietnam) locales, mainly focus on vertical writing test documents.  Some aspects might be due to missing fonts in the test server, this will need to be checked.

Caleb James DeLisle and Aaron McSween (Xwiki, FR) presented their work on collaborative editing with client-side cryptography. Their project Cryptpad promises zero server-side knowledge collaboration, using client-side cryptography to offer defense in depth. This allows for collaborative editing with strong privacy capabilities.

Conclusions

During the first day of the ODF plugfest, dozens of tests were validated by the participants. Egor Goroshko from XCDS/Collab.io held the record with 39 new validated tests. Others focused on the creation of larger text documents and spreadsheets with comprehensive tests on features such as form support. The new tests show a lot of detail previously not available, and uncovered some major interopability issues that were overlooked until now. Some of these have critical consequences for end users.

The key finding is that several major application (Google Docs, Microsoft Office and AbiWord) lose critical information on the internal content hierarchy in documents that are opened and saved by them. The Center for Persons with Disabilities indicates that a uniform heading structure "is often the most important accessibility consideration" in office documents. When encountering a lengthy text document, sighted users often scroll the page quickly and look for big, bold text (headings) to get an idea of its structure and content. Screen reader and other assistive technology users also have the ability to navigate text documents by heading structure. When this information is lost, it significantly raises the cost to process information. The fact that information which is so critical for people with disabilities - and with many other implications - is completely lost upon import is a major bug in any implementation. It also underwrites the importance of systematic testing at the markup level.

Some other interesting results:

• Some implementations seem to treat colors in different and unexpected ways. Possible consequences of this need to be assessed.
  
• Three new tests contributed by Franklin Weng and Ben Martin focused on Asian language support with vertical writing (table-cell-writing mode, paragraph writing mode and section writing mode) revealed errors in left-to-right handling in some applications.
  
• Test findings led to the discovery of some unnecessary redundancies that can be used to further simplify the ODF standard (Thorsten Behrens)
  
• Some logical high level constraints not explained in the prose of the standard or in the schema. One such example is the proper handling of bitmap fill images.

There were a couple of new ideas for server features that emerged from the 12th ODF Plugfest:

• The ability to record who submitted the tests (Michael Meeks)
  
• Possibility to add non-XPATH tests, such as visual tests (Michiel Leenaars)
  
• A comment field per application (Dirk Vollmar)
  
• Possibility to clone a certain existing text (Jos van den Oever)
  
• Handling case-insensitivity.
  
• Quick wiki page with XPath patterns
  
• Button to verify a test for all applications in one go

Also, the recommendation was made that tests that have visual rendering aspects can make sure that these are visible by adding elements to the test that allow comparison. So if a certain font size is set for some element, having more than one element with different (preferably very different) font sizes will make it easier to perform a visual check.

Day II: Thursday November 17th 2017

Location: Paris Open Source Summit

After a welcome by Michiel Leenaars (vice-president OpenDoc Society), dr. Steven Pemberton (CWI, NL) summarized the results of the previous day. After that, a number of presentations took place, highlighting different aspects of the adoption of OpenDocument Format, work done on vertical applications within the European Parliament and European Commission as well as technical considerations. A number of vendors present highlighted updates on ODF aspects since the previous plugfest. All presentations can be found online at: http://odfplugfest.org/2016-paris/programme/ 

Franklin Weng (Software Liberty Association of Taiwan, Taiwan) presented his findings on the m igration protocol used in Taiwan. They suggest "s ide loading" of ODF implementation (in their case: LibreOffice) next to existing solution. Businesses are urged to focus on long term support, not short term selling trainings or licenses. They welcomed Sweden's NPS announcement of 46 IT standards which contains ODF (and excludes partially competing formats such as OOXML). Goverment adoption after thorough research is an important indicator to let people understand that using ODF may be part of a cost-saving policy, but it is more than that. SLAT recommends that in the communication phase it is more important to focus on standardising the format than on which software to use.

Eric Ficheux (Nantes Métropole) presented on the ODF usage in Nantes, where the transition to ODF is complete. They recommend to first remove all the limitations of old legacy format (in this case Microsoft Office 2003 format) before purchasing of new software for organisation and partners. They also showed some guiding principles. Wherever their organisation itself is in control (e.g. its own suppliers, satellite organisations) one can just require ODF usage. Where there is no control (other administrations, partners), they recommend and explain ODF usage. During procurement of new services, the use of ODF is made leading and this has proven a very useful and important choice. Their calculation methods make the long term benefits of ODF tangible: for each bid submitted to a tender, they calculate the true cost they would have with integrating in their environment. Incompatible bids (for instance those that only can work with legacy file formats) are then corrected for this cost: the actual cost of auxiliary software required to use the solution (such as operation system licenses and legacy office suites) is added to the offer. This mechanism helps vendors to adapt to the new requirements by making the cost structure for the customer transparent. Nantes reports great succes, and recommends this strategy to others.  

Géraud Berton (European Commission) talked about LEOS, a tool created by the European Commission for drafting of legal documents in XML. It is an authoriting tool that allows collaborative work, which is a core process of the European Commission. The advantage of having this open source tool is that it is built around the real semantics of the application domain, it can be tailormade to the huge legal workflow inside the EC and can handle the many stages of revisions and review. 

Gianluigi Alari (European Parliament) presented on the e-Parliament Program. This programme by the EP strives for handling structured content for parliamentary texts. This open source project is modernising the European Parliament's core parliamentary information systems. This involves the very parliamentary text production itself, but also the legal verification as well as translation workflows. Amendment drafting in the legacy situation (Microsoft Word) was particularly complex and error prone. The open source authoring tool for amendments (at4am) has made over 400.000 amendments in the EP possible since it was first introduced in 2011. There seem to be some interesting opportunities to integrate with

Egor Goroshko (XCDS, UK/Ukraine) introduced their new implementation Collab.io, built by a multidisciplinary team with experience in a number of solutions such as QuickOffice. In line with the requirement of today they had a multiplatform strategy, and real-time collaboration was part of the original design.  The applications in their suite are built and designed around ODF, which offered some unique advantages: it combined very rich functionality with simplicity, and was easy and safe to extend to cover specific use cases. The team started from scratch, to avoid legacy bugs present in existing code bases. This had another advantage: they could do a single implementation for everything (inspired by ODF), as the design of the standard is coherent in this respect and allows for component reuse. He pointed out that it would be good to have a reference implementation of the standard, this is currently still missing. Egor presented a number of feature requests for ODF, such as: support for themes, paragraph marks, fine-grained track changes, support for collaborative editing. Egor announced that his company will be joining the ODF TC

Michael Stahl (Red Hat, DE) announced that he created a new automated test tool to look for regressions: ODFunDiff. He also indicated two significant bug fixes made since the previous plugfest.

Thorsten Behrens (CIB, DE) presented the work on ODF Change Tracking, where a prototype was built with feature change­tracking. This offers a high level abstraction from ODF XML, and avoids diffing noise. Because it tracks logical / semantically meaningful user changes it applies to different models and allows switching two changes without context. Also, he showed the work on automatic export validation in LibreOffice.

Michael Meeks (Collabora Productivity, UK) gave an update about LibreOffice online and the Collabora CloudSuite.

OpenDoc Society wants to thank all the participants, as well as those persons and organisations that made it possible:

• The French Ministry of Finance
  
• Paris Open Source Summit
  
• Inno3
  
• Forum Standaardisatie from the Netherlands government
  
• NLnet Foundation
  
• Camille Moulin
  
• Laure Patas d'Illiers
  
• Gijs Hillenius