1 # @(#)CHANGES 5.1 (Berkeley) 5/11/93
3 This new version is almost identical to the timed and timedc code
4 that has been shipped for years by a workstation vendor.
6 Among the many changes:
8 improve `timedc msite` to accept a list of hostnames.
10 change slave-masters to answer the packets generated by `timedc msite`
11 with the name of the real master, not their own. This makes it
12 possible to "chase the chain" of slave servers to the ultimate
15 much improve the log caused by `timedc trace on`:
16 -made `timed -t` work.
17 -suppression of repeated entries, which both slowed down the daemon
18 (sometimes catastrophically) and tended to make disks fill up
20 -better time stamps on log entries
22 -dump information about slaves, master, and so on each time
23 a message asking the log be turned on is received, and
24 when the log is turned off.
27 use a hash table to keep track of slaves, instead of the stupid linear
28 list. This becomes handy with hundreds of slaves, instead of
29 the original design limit of "a room with a few VAX's."
31 separate the main protocol timer from that used to look for other networks
34 time stamp packets received by the daemon, so that time corrections
35 are not made (even more) inaccurate by waiting in the internal,
36 timed queue while the daemon is processing other messages.
38 made -n and -i work with subnets not named in /etc/networks
40 compute the median of the measured clocks, instead of the average
43 vastly improve the accuracy of the clock difference measure by
46 use adjtime() when possible, and directly set the clock only when
49 when the requested adjustment is small, perform only part of it, to
50 damp oscillations and improve the long term accuracy of the
53 fix uncounted core-dumps on machines that do not allow dereferencing 0
54 in both the daemon and timedc.
56 fix "master loop detection".
58 fix several cases in which multi-homed masters could get into shouting
59 matches, consuming all available network bandwidth and CPU cycles
60 (which ever runs out first), and convincing all bystanders to stop
61 advancing their own clocks.
63 refuse to behave badly when other machines do. Instead of arguing forever,
64 go off and sulk when other machines refuse to play by the rules.
66 increase the maximum number of clients.
68 add "-F host,host2,..." to "freerun" or "trust" only some hosts. This
69 is handy both when only some machines should be trusted to let
70 root use the `date` command to change time in the network.
72 It is also handy when one machine has some other way of adjusting
73 its clock, whether NTP or a direct radio or atomic connection.
74 "-F localhost" causes `timed` to "trust" only itself.
76 It is also handy to build a hierarchy of timed masters crossing
77 networks. The TSP protocol has no provision of "goodness of clock",
78 no natural way to completely heal network paritions. Judicious
79 use of -F or -G can cause each gateway to trust only itself and
80 machines closer to a central machine with a radio or atomic clock.
82 add #ifdef code that supports NIS "netgroups" of trusted hosts, which
83 can be easier to administer than -F.
85 add #ifdef code to compute an aged total adjustment. This can be used
86 in systems that can make long term changes in their system clock
87 frequency, e.g. "timetrim" in the Silicon Graphics kernel.
90 Problems observed by others that are unresolved include:
92 Practically any users can send to the master TSP messages and this
93 way corrupt the reliability of the system. Authentication
94 of messages should be provided. Unfortunately, that would
95 require changing the protocol with all of the implied
96 compatibility problems. Fortunately, the new -F and -G args
97 can be used to cause the daemon to ignore time changes from
100 MAN. The limit of 1013 on the number of slaves hosts should be doc'ed.
102 It should be dynamically allocated with no limit. On a
103 large network, one host could possibly master over many
104 more than 30 hosts. Given the timers in the code and
105 effectively in the protocol, and the time required by each
106 master to talk to each slave, it is not practical to have
107 more than 200-300 slaves. The master cannot keep up because
108 the slave-chatting is single-threaded. when the master
109 gets behind, slaves start demanding elections. To
110 significantly increase the number of slaves would require
111 multi-treading things, and given that a network with more
112 than 300 directly addressable machines has worse problems
113 than keep the time of day right, not worth worrying about.
115 UGLY,CODE. timedc/cmds.c has a lots of repeated code in it.
117 **** The first thing is that each command is set up as if it
118 were an individual program taking argc and argv. A more
119 conventional calling style should be used. I don't think
120 any of the routines take more than a couple arguments.
122 UGLY. fxn definition syntax does't follow convention:
123 has type on same line.
125 **** It needs to be fixed at least enough that tags
126 will work on it. An entire cleanup might be nice later, but
129 LOBBY(mildly),CODE: Would be very convenient if date(1) took a
130 +-<number> argument to set the time relatively. With
131 the advent of timed it is now reasonable to synchronize
132 with WWV, which is nearly impossible to do "by hand"
133 with just an absolute date, and scripts are too slow.
134 format could be +-nn...nn.ss, where the '.' is required
137 **** If you want to do it go ahead. It sounds useful. As far as
138 syntax goes, the normal format for the date should work just
139 fine for this. If the date is preceded by a plus or minus,
140 the change is relative, otherwise it is absolute.