Berkley DB was not SQL, and accordingly to my own classification of NoSQL (that is, at least one of the two must be true: [1] A different data model compared to SQL, [2] A different tradeoff in CAP compared to traditional DBs) Berkley DB should be classified as a NoSQL database.
However I think it was not a NoSQL database for an important reason, it was only embedded in other programs, and not in the form of a networked server. Not only this raised the barrier to entry, it never made Berkeley DB an "object" in your infrastructure that you could use in different ways, and in a competitive way with other databases.
So the Berkeley DB creators did not started the NoSQL era long time before because they missed the importance of what they were doing, thinking that their DB was limited to just something you could bind to programs not needing all the power of an SQL database. At least this is what emerges form their choices. And the implication is that they were also considering relational databases as the only "real DBs" in my opinion.
So there was no real competition with traditional DBs, and it was not a NoSQL DB.
EDIT: I understand this is a mostly personal point of view, and not an objective critique, but I wanted to share it with HN nevertheless.
And probably the reason is that is wasn't networked running as a server was to reduce network traffic, which was obviously a lot slower (and a bottleneck) back when it was created. Also, it would be trivial to expose its functionality as a networked server, if that's what you needed.
NoSQL is a time period characterized by a set of ideas, not a category. The data stores that came out of NoSQL were about using more specialized tools and rejecting the dogma that all data persistence should be built on an RDBMS.
Besides it's ability to be embedded, BerkeleyDB doesn't really provide any advantages over an RDBMS, other than raw simplicity. It's unlikely anyone looking at data persistence solutions would consider BDB alongside an RDBMS; they solve different problems.
I think BDB is a useful and well-designed tool, but it's not NoSQL.
However I think it was not a NoSQL database for an important reason, it was only embedded in other programs, and not in the form of a networked server. Not only this raised the barrier to entry, it never made Berkeley DB an "object" in your infrastructure that you could use in different ways, and in a competitive way with other databases.
So the Berkeley DB creators did not started the NoSQL era long time before because they missed the importance of what they were doing, thinking that their DB was limited to just something you could bind to programs not needing all the power of an SQL database. At least this is what emerges form their choices. And the implication is that they were also considering relational databases as the only "real DBs" in my opinion.
So there was no real competition with traditional DBs, and it was not a NoSQL DB.
EDIT: I understand this is a mostly personal point of view, and not an objective critique, but I wanted to share it with HN nevertheless.