b6n (Benjman Black) just mentioned a paper on the history of virtual synchrony. This is a protocol for replication, but at the end, the paper makes a few more general observations:

In its early days, the transactional community aggressively embraced diversity. Researchers published on all sorts of niche applications and papers commonly argued for specialized variations on the transactional model. The field was awash in specialized database systems. Yet real success only came only with consolidation around transactions on relational databases: so much investment was focused on the model that the associated technology advanced enormously.

The bold part (my emphasis) reflects the observations I have with regard to ROS: There are certainly more advanced niche solutions out there, but I hope it can focus the efforts of the community.

The paper continues by mentioning that some researchers may have felt the abandoning more specialized solutions as a step backwards -- and this will certainly be true with regard to ROS. However, because ROS is open source, the combined efforts may outweigh that.

To this, Jim Gray is quoted:

In this, Jim saw a general principle. If a technology tries too hard to make every user happy, so much effort is needed to satisfy the 20% with the hardest problems that the system ends up being clumsy and slow. The typical user won’t needed most of its features, and many will opt for a simpler, cheaper solution that’s easier to use. The irony is that in striving to make every user happy, a technology can actually leave the majority unhappy. In the end, an overly ambitious technology merely marginalizes itself.