On Sat, 2006-06-17 at 01:24 +0200, Kurt Roeckx wrote: > I actually have no idea why it was splitted in 2 packages. Afaik > the only difference is that one has all the clocks build in, and > the other doesn't. This results in a difference of the size of > the binary, 200K versus 400K, and I guess a difference in memory > usage. The later seems to be the actual reason why it was split, > and seems to be 450K difference in RAM used.
Yes. Actually, originally there was just an ntp (or xntp, depending on when in history) package. When I turned on all the refclocks, someone complained that the daemon was now large and consumed a lot of memory for the client case... so, I split it into three packages, ntp-simple and ntp-refclock with just the daemon binary, and ntp had the rest. The split of ntp and ntp-server was another step in giving more control over which user space binaries and so forth are present. I personally don't find the 'ntp-server' package name obvious at all. If I'm going to do an apt-get, 'ntp' seems much more natural to me. Don't know that it matters much, but if someone installs 'ntp', it'd be nice if they're left with a system that's time-synced in at least a trivial way. It wouldn't bother me at all if this all went back to just one package, 'ntp', which used a debconf question and either an /etc/defaults/ntp entry or alternatives to control which daemon binary gets run. Keeping two versions of the daemon binary probably makes sense, but any real time lord is going to generate a kernel with the nano patchset and rebuild the daemon against that kernel's nano interface, so the ntp-refclock package is to my mind of only mild interest these days. Hope that helps. Bdale -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
