Re: API, was: Question re HIP dependency & some architectural considerations
Iljitsch van Beijnum <[email protected]>
| Newsgroups | gmane.ietf.multi6 |
|---|---|
| Message-ID | <[email protected]> |
On 29-jul-04, at 15:40, Brian E Carpenter wrote: > "Applications"? The network stack is rarely called by > applications. It's called by network adaptation code > built into "lower middleware" such as the Java run time > system, whose job is to hide all this stuff from actual > applications. Yes, Christian has brought up this point in the past as well. Still, looking up from layer 3 it doesn't matter much whether we have to be compatible with actual applications or Java run time environments and Win32 internals. (There is less of that stuff than actual applications which makes it a bit easier though.) > However, like it or not, multi6 *does* have to deal with > the real world of ULPs that call the current socket API. > That may only be stage 1, but it has to work. Agree. We need to think about stages 1+ too is all I'm saying. > Iljitsch van Beijnum wrote: Yes I know I wrote it earlier today. No need to repeat the whole thing.