Occasionally I get asked why we are relying exclusively on the MarkLogic platform for Epinomy. It would be possible to implement Epinomy on open source platforms, but not nearly as effectively or easily as on MarkLogic. There are several open source components of a MarkLogic-like solution, many of which are garnering lots of attention from the media. We could absolutely cobble together a solution that uses the excellent open source software from Mongo and Apache. But it would have to be a home grown mishmash of dozens of libraries and dependencies just to replicate a part of what MarkLogic does out of the box.
Let's look at some of the features of MarkLogic and how you would implement them using FOSS (free and open source software) tools:
1) NoSQL database
There are several good open source NoSQL databases available on the market. MongoDB is the leader in the space, and is an easy-to-implement noSQL database that runs on Java.
2) Web Application Server
There are all kinds of web application servers available - Jetty, Tomcat, Apache, JBOSS. They each have different features and complexities.
3) Full Text Search Engine
Lucene, Solr and ElasticSearch are all great open source search engines. But they still are not deeply integrated into the database. The capabilities of MarkLogic as a search engine are superior to any open source search engine. Period.
Hadoop provides a good way to store and process gigantic datasets, but it is an infrastructure. MarkLogic can run on the Hadoop infrastructure, but in most cases can function just fine without it. MarkLogic 7 adds dynamic scalability features that are very difficult to achieve using FOSS.
NoSQL databases are not ACID compliant. This means that you can lose data. MarkLogic is fully ACID compliant at a level that no other database can match.
On the worldwide web, security is an afterthought. The general idea is that people can either access the data or they can't. This is not enterprise class or government grade security. MarkLogic has it built-in from the ground up.
There are several open source Triple Stores, but like the search engines, they are not integrated into a complete package, and do not provide the same performance and security as MarkLogic does out of the box.
9) Total Cost of Operation
MarkLogic software can be expensive. But to hire a team that can hand-roll all of the features and integrations necessary to create a platform as robust as MarkLogic is at least as expensive, and probably more so. I will leave it to MarkLogic to do that analysis but I bet it would be surprising for "free" software.
Since all of the components and features are in one package, they are deeply integrated. There are no DLL-hell or Jar version compatibility issues. As their customers say, "it just works".
That being said, we are actively building a Semantic Server on the J2EE platform that will incorporate many of the aforementioned technologies. Java is not dead, and it definitely can do some very powerful things in the enterprise. We also have a .Net product based on SharePoint, so we're not being parochial snobs here. All modern platforms have substantial merit.