The problem I've seen is that every non-relational data engine I've tested has had worse problems than SQL RDBMSs do. Sure, some of them solve some of the problems with SQL DBs, but in every case I've seen, the problems they replace them with are much, much worse.

For example, XML data stores. Yeah, XML translates nicely into screen objects for Web apps. But the performance, etc., I've seen on those are tragically bad. And let's not ignore the issues with transactional integrity that they seem to run into.

For example, Intersystems' Cache, which they promote as a "post-relational database". Yes, its performance and data integrity are excellent. Zero object-relational impedence, since you can directly reference data as a property of an object while at the same time applying joins and such as if it were relational. At the same time, development in it of anything more complex than a phone book database is nightmarishly bad, referential integrity is "we don't need no stinking referential integrity", and just forget tables of lookup values.

Pure OODBs work very nicely with OO applications, but the moment you need to run reports against one or try to do any data mining in it, you can expect a relatively simple project to be something you finish a week before you retire.

I've dealt with a number of these technologies, and the problems I've seen have been at least as bad as RDBMSs, usually much worse.

The day I see something come along that's better than Oracle, SQL Server, DB2, et al, in ways that matter, and not worse in ways that matter more, I'll consider it a good evolution. But I haven't seen such yet.

As for silence on the subject: I see it come up a couple of times a year in various forums, etc. Never gets past the exact phenomenon I'm writing about.