Samisa Abeysinghe wrote: >I think the requirement is to use something other than a file system to >log. >My understanding is that, you can extend the current API to do this. May be >the API semantics are focusing on file based logging. But you can plug your >own impl to do whatever you want. That what I would like to do. >Note that, when you have your own impl, since you are in control, there is >no 'private' scope in the impl. >Re: "add a ''open log' operation to axutil_log_ops_t and move most of the >code in axutil_log_create() to a private axutil_log_impl_open(). " >The concern that I have is how it would impact the current implementation - >because we use this thought the current code - the create and then open >semantics is assumed. And create allocates memory and open opens up >the descriptors. I'm beginning to see the light. >Quite a bit of thought went in to this API when this was >originally designed, and it was the understanding that >this separation empowers extension like the way you want. >The easiest is to make sure you have your own impl, that override >the behavior the way you want. You have to use create, to allow allocation >of memory and intilizations necessary. Then use open the way you want, or >ignore it, if you do not have anything to be done in the space. My error was to try to reuse axutil_log_create() which I thought was the entry point of the API. Instead, I now think have to replace it with my own implementation that returns a struct whose first member is an axutil_log_t. The other data of this struct will be my custom data. Could you confirm? >When you plug Axis2/C logs to your own logging system, you >can initialize logging outside Axis2/C and then plug the writes from Axis2/C >to your logging system via the custom implementation. >I do not see any need to change the current Axis2/C API. I'm begining to understand the design... Olivier.
<<winmail.dat>>
--------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
