[omniORB] omniORB slow transfer when clientCallTimeOutPeriod != 0

Igor Lautar igor.lautar at hermes-softlab.com
Wed Aug 12 13:52:54 BST 2009


On Wednesday 12 August 2009 12:44:22 Duncan Grisby wrote:
> That's not completely true. The Windows bug is more subtle than that.
>
> The slowdown only occurs if the CORBA call arguments / return values can
> be transferred as a single large block, and the block size happens to
> interact with the Windows TCP buffering. If the call data is either
> small, or involves complex types that require marshalling, the slowdown
> doesn't occur.

Thats true, I should have been more clear what triggers the bug (besides non-
blocking sockets and socket send buf size.

Regards,



More information about the omniORB-list mailing list