Interoperability and Best of Breed

There is always a vigorous debate on the strengths and weaknesses of the monolithic integrated approach versus the 'best of breed' interoperable approach.
Health care software is a very diverse and complex field and it is just not possible for any vendor to have high quality offerings in all possible areas. When you consider that medical practices potentially need software that does billing, patient records, disease registry, electronic prescribing, best practice alerts, point of care decisions, laboratory and radiology ordering and reporting, and other functions I think that one can see that the skills involved in developing these functions are unlikely to be strong in all areas.
In addition, as the Nutting report points out, implementation and change is difficult and it is better to take a path of gradual modular implementation rather than try to digest a large application all at once.
Interoperability is difficult because most of our software is not designed for it. The standards exist in HL7, ICD, SNOMED, LOINC, etc but they are not well supported by the integrated software that is currently available.
If you look at the case of the internet standards HTTP, FTP, SMTP, IP, etc. you can see that interoperability works well when you have software that supports the standards. The Internet, web browsing, email, etc all works well using software from thousands of vendors running of a wide variety of platforms.
Unfortunately, medical software has been developed on a closed proprietary, monolithic model with poor support for standards. If physicians understand and demand interoperability, it will appear and vendors who support interoperable systems will prosper.
Unfortunately, the CCHIT seems to be perpetuating the monolithic integrated model of software. Hopefully it can be persuaded to open up to focus on interoperability.