- From: John Cwikla <cwikla@wri.com>
- Date: Fri, 12 Jan 1996 11:58:57 -0600 (CST)
- To: frystyk@w3.org
- Cc: gunnarr@ifi.uio.no, tag@hpcvlx.cv.hp.com, www-lib@w3.org
-> The event model is currently built so that is possible. When a socket blocks -> in the middle of a request, the load module calls the event module with the -> socket in question and the request object. These two are then be registered -> together so that the request object associated with a socket can be found -> again. -> -> However, in some situations you have more than one socket pr request and/or a -> socket can last for more than one socket. This is for example the case with -> FTP and HTTP using persistent connections. The HTNet object is used to handle -> this and also to be the glue between the socket descriptor and the request -> object. The HTNet object represents a "library thread" as it has a socket -> _and_ and request object associated with it. -> -> It also has a pointer to the right callback function (that is, the load -> module) so that when a socket is ready for action, both the request object, -> the load module and the socket descriptor can be found again and the request -> continued. The relations between the objects look something like this: -> [ snip...] -> I think that the method above will result in a more integrated application -> with less glue code. -> -> -> -- -> -> Henrik Frystyk Nielsen, <frystyk@w3.org> -> World-Wide Web Consortium, MIT/LCS NE43-356 -> 545 Technology Square, Cambridge MA 02139, USA Actually this isn't what we want. We don't want to have an event loop out of our control, or one isn't the event loop of our main app. This forces the WWW loop to have control and to come up with some hack to solicit events (say from a GUI) or to respond to events. What we want is to be given a file descriptor, a flag, something, that the event loop in OUR program can then solicit and decide what to do. We can't have some library, that is likely to be a module in a larger program controlling event dispatching... -- o John L. Cwikla o o o X Programmer o I'm not in a bad mood. o o Wolfram Research, Inc. o I'm always like this. o o cwikla@wri.com (217) 398-0700 o o o http://www.wri.com/~cwikla o o o http://www.wri.com/~cwikla/widget o The Widget FAQ o o http://www.wri.com/~cwikla/xlopedia o The Xlopedia o
Received on Friday, 12 January 1996 13:00:14 UTC