my point is ,you dont write raw tcp text commands when you do an http request in Go(well you can,but what's the point when... )...There is an abstraction layer,you deal with "objects",not HTTP text commands directly.
I dont see why it's nice to have an abstraction for HTTP but as soon as it comes to RDBMSs, string commands should be the way to go...
Your analogy is flawed. Using a httplib for http is like using a database driver for SQL. An ORM for http would be something like Ruby's Active Resource or some other abstraction that his all URLs and query strings from you.
Because it's a massive impedance mismatch for most nontrivial cases. Relational and object systems have never played particularly well together. The contention is, and it's one that I agree with, is that the tooling that inevitably builds up to address this impedance mismatch is too hard to usefully generalize in a way that doesn't end up taking over your project.
(This is why I dig mappers for document-based databases and avoid them for relational ones.)
I dont see why it's nice to have an abstraction for HTTP but as soon as it comes to RDBMSs, string commands should be the way to go...