[ntp:hackers] Multicast

David L. Mills mills at udel.edu
Sun Oct 2 19:37:54 UTC 2005


Danny and Tribe,

I can wait, but I don't think ntp-dev should be released unless 
broadcast and multicast works in IPv4 and IPv6.

By the way, the compiler problem was wierd. Turns out the error report I 
found has been there for some time and apparently nonfatal and probablyl 
inconsequential. My problem was a spurious character before the first /* 
comment. Everything works now except make dist, so I can't move to 
campus and test there.

Note to all:

1. I hate it when some install mucks up my man directories. Assuming 
there are man pages in the distribution, of which I heartily disapprove, 
there should be a make man target. I would much rather the man 
documentation be a separate distribution.

2. I took some effort to refine and produce the html documentation for 
the original sntp program. If the man page was automatically produced 
from that documentation and would be changed if I changed the source, no 
harm is done. Otherwise, I will discontinue the html documentation for 
that program.

3. I looked at the sntp man page and got stewed all over again. It 
leaves my X window trashed with underlining, reverse video and general 
chaos, even after the man page is closed. On the other hand, Linux does 
that, and juggles the colors, too. FreeBSD and Solaris take some 
liberties, but at least leave my window as it found it. Can we entirely 
resist underlining, reverse video and anything else for emphasis, or at 
least leave the window as found? The underlining in particular is really 
hard for me to read and the reverse video is startling.

4. If the sntp man page continues, users would be mightily surprised if 
there is no man page for ntpd, which could at the least point to the 
html documentation. I still plan to remove all the build pages from my 
master pages on the assumption this information is provided not by me.

5. The sntp program has nothing to do with the ntpd daemon and support 
programs. It and accompanying documentation should be a separate, 
independent distribution, as should many of the scripts and other stuff 
unrelated to the ntpd function itself. Some of this stuff either doesn't 
work or has been deprecated (like my 10-year old log scripts in a now 
deprecated language). There should be a contrib distribution with this 
excess stuff.

6. I don't believe the sntp program should be released until the NTP WG 
sees it and verifies it is compliant with the SNTP rfc now in the pipe. 
I want to see very strong words that reassure folks that the program is 
squeaky clean, adheres to the spec chapter and verse and passes what 
portions of the test suite that apply to it. This includes setting the 
clock forward and back throughout NTP era 0. This is exactly what 
tripped up ntpd in 2004, now fixed. Does it verify the originate 
timestamp as suggested in the rfc? Does it do any of the other tests 
TEST1 through TEST3 as ntpd? It's very easy to do this and it does 
distance it from the other knockoffs now available. Does it do anything 
with KoD messages, like display the kiss code? That would be very useful 
for the newbie and probably overlooked by the knockoffs.

Dave

Danny Mayer wrote:

> David L. Mills wrote:
>
>> Guys,
>>
>> Add to my last. Broadcast works, but not multicast. When adding 
>> 239.1.1.1 to the list of addresses an error message 
>> setsocket_IP_MEMBERSHIP results
>>
> Dave,
>
> That doesn't surprise me based on your earlier message. I will have to 
> look to see what's going on, but it not likely to happen in the next 
> few days.
>
> Danny




More information about the hackers mailing list