On Mon, 15 Mar 2004 13:22:14 -0800
"David S. Miller" <davem@xxxxxxxxxx> wrote:
> On Mon, 15 Mar 2004 09:20:18 -0800
> Stephen Hemminger <shemminger@xxxxxxxx> wrote:
>
> > > Would it be too complicated, if sysctl would give the global default,
> > > from which one can deviate using RTAX_FEATURE?
> >
> > Maybe for frto it needs to stay, but sysctl's are more painful and complex
> > than keeping the stuff in the routing info. Also, the external tools are
> > part of every distro, except for a few embedded systems, the networking code
> > depends on user tools already.
>
> How do you propose to support some kind of "global enable" for features.
The easiest way to do that is to initialize each TP with features from
sysctl when created.
> I think sysctl's support this quite well. The test for the feature
> becomes "sysctl || route_attribute".
That is what the next version does for FRTO.
> Also, as Yoshfuji stated, you absolutely cannot change the existing
> sysctl numbers as tools that use the sysctl() system call use those
> numbers explicitly thus they are compiled into applications.
Okay, what about WESTWOOD?
> I really am not going to consider something that removes existing
> sysctl tunables. :-)
What about tcp_westwood which is new?
|