Datasets and XML in place of proper db: Not a good idea?

Question

In continuation of: http://stackoverflow.com/questions/729634/storing-datarelation-in-xml

Thanks to everybody for answers to my earlier thread. However, could I ask the reason why everybody is not supporting this XML based approach? What exactly will be the problems? I can apply connstraints to dataset, and I can, I guess, also use transactions.

I am new to this. So if you could point me to some link, where I can find some sort of comparisons, that would be really helpful.

According to FAQ, discussions are not very encouraged, but I guess this is quite specific. I hope, not to be fired for this... :)

Thanks for reading, Saurabh.

Answer 1

Concurrency will be the main issue, where multiple users want to access the same "database" file. Performance is the other, because the whole file has to be loaded into memory. If the filesize grows it'll get unmanageable. Also, performance on queries is hit, because it won't be as efficient as getting something as tuned and honed as an RDBMS to do it for you.

Answer 2

Though I would need to research to back this up, I suspect that there are some performance implications to not using an actual database.

If there is a chance that another application built on a different platform (not ADO.NET) might need to access your data in the future, having to work with a giant XML file will very likely make life more difficult. A relational DB is the standard approach to this sort of problem.

Answer 3

Database management systems are specifically designed to store data and retrieve it quickly, to preserve the integrity of the data and to leverage concurrent access to the data.

XML, on the other hand, was originally designed for documents, separating the content from the presentation. It became a handy way to store simple data because the file structure is so well defined, and then it went out of hand with people trying to store entire databases in an unsuited structure.

XML doesn't guarantee atomicity, concurrency, integrity, fast access or anything like that. Not inherently, anyway. .NET's DataSet libraries do help in that regard, but just because you can serialize DataSet objects to XML doesn't make it a good place to store data for multiple users.

When you're faced with two tools, one which was designed to do exactly what you need to do (in this case a DBMS) and one that was designed to do something else but has been kludged to do what you want, sorta (in this case XML), you should probably go with the first option.