Today I sent the finalized DRMAA 1.0 IDL spec to the mailing list for review. Before this, I tried to compile the resulting IDL in some IDL compiler. Together with Martin and Harald, we tried omniidl (from OmniOrb 4), fnidl (from the Python fnorb), idlj (from JDK 1.5), Quedo’s IDL compiler and JacOrb’s IDL compiler. To my very surprise, all of them failed with a correct IDL file ! Specifically, all of them failed for a getraises statement in one of the attribute declarations. It seems like there is no widespread support for the new IDL concepts from CCM / Corba 3.0. fnidl also failed with the valuetype declarations, which might render it a little but useless for non-trivial examples. Quedo needed some hacks in order to work, and JacORB wrongly claimed a naming clash. omniidl and idlj are now happy after removal of the getraises statement.

Please note that the DRMAA IDL spec is not intended to provide an input for an IDL compiler. I performed this experiments only to make sure that my IDL definitions are syntactically correct. Language binding authors should map the IDL keywords to their best-suited language constructs, as explained in the introduction chapter.