> On Sun, 25 Jan 2026 16:17:49 -0500 > [email protected] wrote: > >> Classic databases are a poor choice for arbitrary data. They are not >> designed for that work flow. That is irrefutable. There are some very >> good object databases designed for that work flow, rocksdb by >> facebook (based on LevelDB). > > Even so, object databases have vertical scaling limits for very large > objects, for some value of "very". The RocksDB FAQ says that keys and > values should not be more than 8MB and 3GB respectively. Is 3GB very > large? That depends on your data. It might be sufficient for email > messages... until someone starts mailing "Linux ISOs", blows out the > database and crashes your mail system.
Yes, that's why I said file systems are designed for data storage and retrieval. Classic databases are design to keep data integrity and provide an algebra to "find" the data you want. Things like, for a lack of a better term, object storage systems, canonically called object databases, tend to blur the line. Classic databases have a rigid definition for their data and arbitrary data either needs to be formatted to fit the architecture of the database *OR* the database has internal modification to manage it. Its all just pros and cons in accounting, but you should really understand the systems, and how they work, when evaluating how to architect your project or product. > > -- > \m/ (--) \m/ > _______________________________________________ > Discuss mailing list > [email protected] > https://lists.blu.org/mailman/listinfo/discuss > _______________________________________________ Discuss mailing list [email protected] https://lists.blu.org/mailman/listinfo/discuss
