Oliver,

I'm going through your remarks, implementing those that we agree.

Yet, I'm not sure of 1 issue:

On Wed, Feb 09, 2011 at 08:31:06PM +0100, Oliver Hartkopp wrote:
> On 03.02.2011 10:39, Kurt Van Dijck wrote:
> Dito. Here you already detected the problem by yourself :-)
> 
> > +  - /proc/sys/net/j1939/tp/retry_ms [int]:
> > +    Controls how many time to wait before retrying to send an individual TP
> > +    flow or data packet after transmission failure.
> 
> Make it a constant that can be overridden via sockopt.
> 

This (and maybe others, but let's stick to this one) property helps shaping
the capabilities of the j1939 stack. Its goal is to help to get a balance
between flexibility, performance, ... versus memory & cpu consumption.
They are not only determined by the platform, but also on the bus that the 
system
is connected to.

Using ethernet/ip as equivalent, I see similarities with:
$ ip link set eth0 promisc BOOL
$ ip link set eth0 mtu VAL
which are link properties that help the system administrator.

These properties are now (for j1939) in /proc, since I just recently
implemented netlink.

I can understand not using /proc for these, but I still think these are
'link specific' properties (now they're global), and not socket specific.

example:
retry_ms : when replying an RTS with a CTS frame in a transport session,
it's not clear yet _if_ a socket will be interested in the resulting packet
when it's complete. If a retry is needed, the delay is IMO link specific.
A constant is a bit rude I think, but I don't think this is a heavily used
metric.

Kurt
_______________________________________________
Socketcan-core mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-core

Reply via email to