|
|
|
|
|
|
by swingline-747
2819 days ago
|
|
|
No, that's an misinformed, all-or-nothing, gross-mischaracterization. Data management systems are a toolbox. The core tenent, distilling ACID, of RDBMSes is referential integrity: that what a foreign key points to can't simply disappear and leave a dangling reference. If you have a database of customers and their orders, you want orders destroyed with that customer, and the referring customer can't simply vanish. There are several delete dependent policies, depending on the use-case. Key-Value stores (NoSQL) typically make no such guarantees. KVSes and Document store are fantasic for less critical and more nondeterministic data like user profiles, likes and tags... things that don't matter as much, as say, bank accounts transactions. Only a fool would use eventual consistency where it could be gamed by making maximum simulateneous withdrawls of say a $1000 at multiple locations. This is where atomicity and consistency are very important. Use the right underlying data guarantees for the task at hand... not like a consulting company whom tried to turn a DBMS into a NoSQL by having a single database table store everything. IIRC the columns looked like this: Key | Value | Type | Notes
Good luck indexing that pile of slow, and say hello to full-table scans for nearly every fetch. |
|
|