In order to understand different implementation altrenatives I compared each of the alternatives that I could see using the constraints that I believe are important. This is what I came up with:
There seems to be the following alternative data storage options for implementing an XBRL based system:
- File system: Store information as files in your file system.
- Relational database: Put the data into a relational database.
- Multidimensional database: Put the information into a multidimensional database.
- RDF triple store: Put the information into an RDF triple store.
- NoSQL database: Put the information into a NoSQL database.
There appear to be two different storage strategies which one might use:
- XBRL technical syntax format: Store information in the raw XBRL technical syntax format.
- Queriable semantic oriented format: Convert the XBRL technical syntax into a more queryable semantic oriented format so that you don't need to convert the information at query time, thus hindering performance.
These seem to be the important evaluation criteria:
- Easy for business user to use (intuitive): Is the system EASY or HARD for a business user to use?
- Query power and query sophistication: Is the query language POWERFUL or UNSOPHISTOCATED?
- Performance, query speed: Is query performance FAST or SLOW?
- Expressive power: Is the system EXPRESSIVE or INEXPRESSIVE?
- System flexibility, agility: Is the system FEXIBLE or INFLEXIBLE?
- System scalability: Is the system SCALABLE or NOT SCALABLE?
- Global standard: Is the system STANDARD or PROPRIETARY?
- Cost effective: Is the system INEXPENSIVE or EXPENSIVE?
- Maintainability: Is the system EASY TO MAINTAIN or HARD TO MAINTAIN?
So, I created radar charts for each of the options for implementation from above. On the chart I show two sets of data: one for storing the XBRL technical syntax (in RED), the other for storing a more business semantics oriented format (in BLACK). I assigned values for each of the 9 constraints. The radar graphs below show the results that I get:
File system
(Click for larger image)
Relational database
(Click for larger image)
Multidimensional database
(Click for larger image)
RDF Triple Store
(Click for larger image)
NoSQL Database
(Click for larger image)
This is the raw data that I used to populate the radar charts:
(Click image for larger view)
If you don't like the values I assigned for each alternative for each of the constraints, here is the Excel spreadsheet which I used, assign your own values.
You can download the radar graphs and data which are summarized in this PDF.
Reach your own conclusions.