On Tue, May 22, 2001 at 05:27:21PM -0700, [EMAIL PROTECTED] wrote:
> On Tue, 22 May 2001, Roy T. Fielding wrote:
>
> > > WIBNIF we reintroduced the notion of a network layer IOL? For the above
> > > mentioned SSL API, we would insert the appropriate IOL during the
> > > pre_connection phase. The core filter would use iol_read, iol_write, etc.
> > > to do the network i/o. Thoughts?
> >
> > How is that different from swapping the bottom filter (doing normal write,
> > writev, sendfile) with an SSL filter (doing ssl_write)?
>
> The main difference is that currently conn_rec's have an apr_socket_t in
> them. If we really wanted to abstract this just using filters, that would
> need to move to a void *, so that modules could store arbitrary types of
> structures in it.
Well... the MPM kind of needs an apr_socket_t somewhere, so it can do its
work. But yah... I could see that field in the conn_rec being left as NULL
in case the MPM has a connection that didn't come over a "socket"
Note that modules would simply use conn_config to store private data. We
don't really need to alloc a field for them.
And filters would use their context field, rather than conn_config.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
