Azure differences between local fabric and cloud handling of datatypes
It is not possible to make local storage and fabric behave exactly like the production counterparts. There are completely different implementations.
The practice that we've adopted at Lokad based on our experience with Azure is:
- start using production storage and fabric as early as possible in the prototyping and development;
- if really needed, build development abstractions and helper classes, that hide or handle critical differences or sharp edges of Azure.
Some of these abstractions and classes are made available in the open source projects for Azure:
- Lokad.Cloud (storage mapping and distributed executor)
- Lokad.CQRS (CQRS-style scalable architectures for Azure)
I think this is the page you want: http://msdn.microsoft.com/en-us/library/dd320275.aspx (It does mention the discrepancy you saw.)