On Mon, Nov 09, 2009 at 01:58:03PM +0100, Thomas Jansen wrote: > On Mon, Nov 09, 2009 at 01:34:16PM +0100, Diego Biurrun wrote: > > > + * Initialize a new packet. This has to be the first function called > > > after > > > + * hip_msg_alloc as we reset the packet size to indicate that no > > > parameter has > > > + * been added yet. > > > > .. which makes me wonder why hip_msg_alloc does not take care of this .. > > hip_msg_alloc just allocates a buffer and memsets it to 0. hip_build_user_hdr > takes such a generic buffer and initializes the HIP packet within. It's not > necessary with our current use of those functions to separate both steps. > However the separation is quite useful if you want to receive such a message > and just need a buffer. As you read the packet from the socket the basic HIP > packet layout created by hip_build_user_hdr would be overwritten. > > If the implementation of the HIPL interface remains this specialized we can > merge these functions, if someone were to pick it up from here and extend > this stub to be a complete library the separation might come in handy. I would think that HIPL itself rather than PISA would serve as the basis for such an endeavour. Thus, if a merger simplifies the usage of these functions, I think we would benefit from it. Diego