A key program point of the 2nd International Workshop on Semantic Web Enterprise Adoption and Best Practice (WaSABi 2014) was the breakout brainstorming session, at which challenges regarding enterprise adoption of Semantic Web technologies were discussed, and potential solutions to some of those challenges described.
The requested output of this session was for each group of 5 people to produce a few slides detailing such challenges and solutions, with an eye towards possible future collaboration areas between the academic and industry representatives.
Each group was provided with conversation starter sheets on enterprise adoption topics identi ed by the workshop chairs, and given an hour of time to discuss and summarise issues related to these topics. The resulting slides are attached to the WaSABi 2014 proceedings. Additionally, the contents of the presentations accompanying the slides are summarised in the following sections. One group discussed topics related to maintainability of systems based on Semantic Web technology. This is obviously a fundamental software quality issue that needs to be handled if enterprise is to adapt such technologies; systems must be maintainable for the duration of their lifetime, which may be signi cantly longer than the lifetimes that academics developing prototypes for research papers are accustomed to. Three challenges in particular were considered, though this list is far from exhaustive: { The longevity of URIs in a changing environment - the distributed nature of the Semantic Web and the Linked Open Data cloud means that companies may make themselves dependant upon resources that are outside of their sphere of in uence. What might happen if an entire critical namespace were to go o ine due to lack of funding or interest? For instance, if xmlns.com (which hosts the FOAF namespac) were to shut down, or purl.org, or schema.org, etc? { Persistence of vocabularies/datasets - Related to the above, what might happen if a resource published on a particular namespace were to be modi ed in a way that is incompatible with existing systems that makes use of said resource? { Communication of popular technologies and vocabularies - A lot of ontologies and technology are released by academia, but do not reach su cient degrees of adoption to warrant maintaining in the long term. How can enterprises (and academics!) measure the adoption rate and subsequent chance of longterm survivability of such technologies and ontologies?
Another group discussed the issue of hiding complexity from software developers and simply making the tech work. For Semantic Web technology to be adopted, it is crucial that the learning curve be as at as possible. Otherwise established methods and technologies, that may be poorer in a number of perspectives than their Semantic Web counterparts but are still su cient to solve the task satisfactorily, will win out. Consequently, the Semantic Web community needs to consider how the complexity of the Semantic Web stack can be abstracted away and hidden from software developers. Here some issues were particularly singled out: { What are the reference solutions / best practices concerning some common and basic developer problems? - Some examples where there are presently no single leading Semantic Web technology include: ORM/DAO transformation, ETL, and LDP/Middleware. { Where to make the cut skipping Semantic Web internals? - That is, how much Semantic Web technology do we expose to the developer users, and how much do we abstract away? If we hide too much of the complex stu , there is a risk that developers to not make full use of the power of Semantic Web technologies, but rather consider it \yet another data storage mechanism" or \just another API". { Scalability: Volume, Velocity, Variety - There is in the research community still signi cant discussion on how to reach scalability using Semantic Web technologies, whether general purpose SPARQL endpoints are usable at all, whether REST-ful APIs are the way to go, whether Linked Data Fragments may be a way forward, etc. Developers do not want to deal with this uncertainty, nor having to make hard judgements about tradeo s and technology ahead of getting started with a project; they just want tech that works sufciently well and with the promise of future scalability. 3
Regarding the rst set of problems, the maintainability of systems based on Semantic Web technology, a number of potential solutions were suggested. Again, this list of course not exhaustive, but may be a useful starting point. 3.1
Developers building vocabularies that they expect to be around for some time, or to be used in a non-trivial context, should seriously consider under which namespace they publish that vocabulary. It is likely that using the web hosting provided by their current employer is a bad choice, as a simple change of jobs might break the namespace. The suggestion of the breakout group is to make use of the W3C Vocabulary Services7, available to any W3C Community Group (i.e. you'll need at least ve people accepting to be part of the community maintaining this namespace). The breakout group has a hard time seeing any organisation more likely to maintain a consistent and stable long-term namespace than the W3C.
When it comes to existing vocabularies that are hosted on namespaces that are not guaranteed to persist in the long term, the breakout group suggests that the vocabulary authors/hosts analyse the access logs of said vocabularies to ascertain the number of users that actually make use of these namespaces, before performing any changes to them. In the case that these resources are extensively used by the community (e.g., FOAF, schema.org, etc) it may be argued that such statistics be of public interest and ought to be released to the community if possible. This might also indicate adoption rates, which can help enterprises make technology choices.
Another idea that was brought up was studying whether the traditional DNSbased URIs are in fact necessary in the future. Considering that there are technologies such as BitTorrent, which uses DNS-less anchor links for resource resolution, possible similar technologies could be repurposed for the Semantic Web? The breakout group is not very knowledgeable about the internals of BitTorrent technology, but suggests that these types of technologies and solutions be studied also. 3.2
The rst recommendation of the breakout group concerning this topic was vocabularies and datasets, once they have been published on a publically resolvable URI, should NOT be changed. This is in line with the existing W3C Best Practice Recipes for Publishing RDF Vocabularies8.
Additionally, the breakout group suggested that easy to understand and easy to interpret change logs between di erent versions of a particular ontology (published on di erent namespaces) would be very helpful, to let developers know whether to adapt their systems and dependant ontologies to newer versions of some ontology. As far as the breakout group knows, there is presently no standardised manner of displaying such deltas in a developer-friendly manner, and this would be a valued research contribution. Related to the same issue, there are a number of annotations in OWL supporting versioning, but these are only 7 http://www.w3.org/2013/04/vocabs/ 8 http://www.w3.org/TR/swbp-vocab-pub/ sparingly used. The group recommends that vocabulary authors learn about and use these annotations to a greater degree.
A further suggestion for future applied research (possibly a MSc project?) was the development of methods and tools to analyze a code base (primarily in Java, as this is the dominant language for Semantic Web development) and produce reports on the technology and resource dependencies of that code base. Such a tool could help developers nd dependencies such as vocabularies, software libraries, etc. that they might want to back up or clone into their own version management systems. 3.3
Communication of popular technologies and vocabularies The breakout group suggests the use of LOV9 and LODStats10, which are concerned with exactly this type of work for vocabularies/ontologies.
Concerning technologies and solutions, no equivalent exists. One suggestion was to consider social networking ideas; possibly setting up a community site with a high degree of interaction among users, listing technologies, and providing information about the pros and cons of each, known installations, etc. Presently semantics are discussed on many di erent places on the web; GitHub, SemanticWeb.org, answers.semanticweb.com, the ONTOLOG community, etc. Perhaps integrating such content into one place, and combining this with a list of well known technologies and installations, could be a useful addition to the community. 4
The group suggests that Semantic Web researchers need to become more familiar and comfortable with composing systems using abstraction layers that hide functionality. This necessitates a degree of quality assurance not historically associated with research software, as those abstractions layers, i.e. APIs, will need remain stable and supported over time. The Semantic Web research community has a lot to learn from our practitioner partners, with regards to testing, packaging, and maintaining software!
In terms of practical development, the group suggests employing REST-style architectures, which the practitioner web developer community are already familiar with. Standardising on this type of architecture also allows our software to more easily become interoperable with non-semantics based components and systems, to the bene t of both sides.
The group notes that research in Semantic Web technology is becoming more driven by practical applicability, as exempli ed by technologies such as JSONLD (supporting the use of semantics with existing toolsets), Turtle (supporting RDF that human developers can read), Linked Data Fragments (supporting less 9 http://lov.okfn.org/ 10 http://stats.lod2.eu/ expressive but very scalable and e cient querying), and schema.org (supporting simple data content schemas that search engines understand). These technologies, and many others, trade a little bit of technical or scienti c \purity" for everyday usability by web developers: in the opinion of the group a very worthy tradeo to make. Some of these developments are being lead exclusively by companies; others by academics with an understanding of enterprise needs. The continuation of the WaSABi workshop series seems highly important for supporting both categories of developers.