On Friday 14 November 2003 11:28 pm, Jonathan Stone wrote:
> (Did the fast-ipsec commits yesterday have anything to do with this
> resurfacing?)
>
> I think we *did* decide, in that kre's articulation is sufficient to
> convince any reasonable and knowledgeable person.
>
> I'm willing to either remove the extant ip_randomid code altogether,
> or [if preferable] replace it with a call to a ``allocate an ip-id
> with whatever policy is currently in force'', together with a knob to
> select either the old-style increment-a-global or the current random-ip
> code. (knob can be config-time or sysctl, whatever is desired).
>
>
> That at least decouples the issue of the low quality ip_ids we get
> today in -current, from the issue reworking the extant random-id code
> or replacing it with something better.

No, what it does is maintain a piece of code that does not work correctly or 
as advertised.  It should be removed.