Mozilla announced an open source version of the Netscape client far in advance of feasibility, especially when including the mail/news client was concerned.
The 4.5 through 4.7 versions of the client, were hopelessly entwined with a third party object oriented database.
The need was to invent an abstract interface that could be satisfied by more than one database backend, without unduly favoring any particular style of database. This is where the MDB interface came from.
The idea was to push the third-party OO database in C++ under this interface for private Netscape versions of the client.
Then some open source replacement could be used in open Mozilla versions of the client, without disclosing anything about the third party database.
The order was to not use a "real" database in the open source client for Mozilla, but to use something plain text based.
Mork was invented out of whole cloth to satisfy requirements for flexibility, resistance to corruption, and brevity in file format for expected content. The last issue about size was the one that mattered most to Netscape management.
They could have ordered the use of XML, but didn't. Main concern was about user data loss.
There was a chance that evolving XML implementations could cause interesting quirks in data availability; the code would be immune to entropy caused by fads and fashion.
Losing data as Mozilla evolved would have been a black eye in public relations wars.