Subject: Re: Wow. (Was benchmarks...)
To: None <current-users@NetBSD.ORG>
From: George Michaelson <G.Michaelson@cc.uq.oz.au>
Date: 12/06/1994 15:25:15
>> I added tickadj to /kern. Someone (cgd, I think) said it belonged
>> in sysctl, which I can sort of understand, but can't understand why
>> that would be cause to be uninterested in having it in kernfs.
> 1) I already put it in sysctl, but I explicitly did *not* make it
Then as far as my need goes you might as well not have bothered; the
reason I put it there was so I could change it at runtime.
You've never justified *why* it's a good idea to patch it at runtime.
Is there some reason you can't change it in param.c? If the default
is fundamentally wrong, then perhaps it should be changed globally.
My understanding is that NTP V3 believes it can analyze the clockdrift
across very long baselines, and derive an "optimal" value for this which
can then be runtime stuffed back in at reboot or on demand.
Coupled to the drift file this means more accurate clock. In fact, I
think ntp provides a tickadj tweaker for just that purpose.
I guess you can rebutt that if its long-lived enough to be specified by
a kernel tweak in /etc/rc.local then it can be a compile-time param but
there might be some utility in being able to frob this without a make...
I won't complain either way if there's a sane documented way to make a NetBSD
box do reliable NTP time sourcing.