Another discussion I watched recently was Dan Pritchett on Architecture at eBay
where he mentioned that much of what they are doing at eBay
is discovering that there may be better data stores for certain kinds of data when building extremely high volume sites.
The combination of both podcasts made me wonder how the other data stores (not likely relational) used at eBay
deal with the object impedance mismatch. The other thing that occurred to me is that it would appear that a tool like nHibernate
may be better positioned to deal with differing datastores than tools like Linq to SQL
as it is provider driven rather than tightly coupled to its store implementation (Linq to XML
, Linq to Entities
to <wait until Microsoft produces one that matches up to the item you are looking to Linq
to>). I think that the pluggable architecture of WCF
corrected this issue and Linq
may have missed it.
I am not saying that nHibernate is the answer (see my previous post on nHibernate) but it would seem to be better positioned than Linq to work with what could be a new way of looking at application state persistence if we consider how very large firms (the eBay tie in) utilize data stores.