Re: Zombie request-route with 2.0.31`

Olaf Titz (olaf@bigred.inka.de)
Tue, 11 Nov 1997 11:09:50 +0100


> > We've had this time and time again: there is currently no other
> > mechanism which can completely replace request-route and there are
> > applications that depend on it.
>
> Well thats a pity, because we removed it.

...based on an invalid premise (that it would not be necessary).

We do still have the old stat() call with the small fields in the
kernel in order to not break compatibility with 1992 libraries, but
you insist on breaking a feature which is needed _now_ for the sole
reason you don't like it.

Yes I know the "this can be called from interrupt" argument, yet noone
has identified the code path where the message sending could sleep,
allocate memory etc. The people who designed kerneld were not stupid
and put some special-casing in to take care of that.

> Diald is just an example of how to do it, and the same technique holds
> for other applications

Not for non-static routing.

olaf