James Carlson wrote:
> I don't think I understand that.
> 
> In order to issue the warning message, we would have to detect the
> case in which someone is setting a destination address, but should not
> be doing so.
> 
> If we know enough about the exact case in which this occurs, why
> wouldn't we just return an error and prohibit the action instead of
> emitting a warning message?
> 
> The answer has to be that we don't know the difference between the
> cases where it happens and it's wrong and those where it's actually
> right.  If we emit a warning message if the command is used
> _properly_, wouldn't that just introduce a new bug?  (And likely also
> customer calls from people baffled at what's broken ... ?)
> 

By way of mitigation, the last message was written under the assumption 
that this issue falls under the category of "not worth the bother and risk."

I completely agree that the problem should be fixed rather than ignored, 
but if the risk is to great to fix it outright, why not issue a warning 
in the meantime? At the least, it notifies a user (or engineer) that how 
they make use of pointopoint interfaces may need to be re-thought out.

It could also be that I am oversimplifying the problem; it would seem 
that if IFF_POINTOPOINT is not set on the interface at the time of the 
SIOCSIFDSTADDR ioctl call, a warning would be issued via cmn_err (the 
call would still succeed as it does today).

I suppose there is also the option of simply categorizing this as a 
feature and not a bug.

-- 
Yet magic and hierarchy
arise from the same source,
and this source has a null pointer.

Reference the NULL within NULL,
it is the gateway to all wizardry.
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to