"mirror" regets whole Debian archive
I had for 6 months used "mirror" to mirror the Debian distribution through
the site uiarchive.cso.uiuc.edu.
On August 30, 1997, MY MIRRORING PROCEEDED TO GET THE WHOLE MIRROR,
REPLACING FORMER IDENTICAL FILES.
I did not change mirror sites, I did not change my mirror.defaults file,
though I did turn my computer on after a 3 week vacation.
This regetting-the-whole Debian archive also happened to me last year
under different circumstances when I made many changes in my computer's clock
setting, my debian mirror site, ... .
This is a bit inconvenient with my modem,
requiring a week of constant connection to my ISP.
Checking some dates as my "mirror" replaces files,
I present three example files [my computer has Eastern Time in the US]:
file date before date now date at
ftp.debian.org date at debian.crosslink.net
and at uiarchive
[on Eastern Time in US]
-------------------- ------------ -----------
---------------------- ----------------------------
perl_5.003.07-10.deb Apr 22 5:24 Apr 22 13:24 Apr 22 8:24
(5 hour) Apr 22 12:24
perl-debug_5.003.07-10.deb Apr 22 5:23 Apr 22 13:23 Apr 22 8:23
(5 hour) Apr 22 12:23
ucbmpeg_lr2-2.deb Mar 20 19:52 Mar 21 4:52 Mar 20 22:52
(6 hour) Mar 21 03:52
Either uiarchive.cso.uiuc.edu has added 8 hours to its archive dates on files,
or my computer has somehow changed its dates.
My computer only adjusts system-time via ntpdate or xntpd,
and has without-problem made these changes for some months now.
I just returned from a vacation of 3 weeks.
QUESTION 1: When "xntpd" and "ntpdate" change Linux system-time,
or when "clock" changes the computer's CMOS time,
do the filename dates change?
Could something horrible (nothing I have seen) have happened to
my disk-drive/scsi-bus that changed file dates?
I ran some experimental tests, mirroring the file
debian/bo/binary-i386/hamradio/rspfd_0.04-1.deb from various sites.
The uiarchive.cso.uiuc.edu version was replaced
as follows (ordered by longitude):
ftp site location file date result
------- -------- --------- -----
ftp.debian.org.au Australia Mar 20 14:52 no change
ftp.uwa.edu.au Australia Apr 16 15:14 replaced
uiarchive [obviously the April date is wrong]
ftp.riken.go.jp Japan Mar 20 13:52 no change
sunsite.doc.ic.ac.uk UK Mar 20 22:52 no change
uiarchive.cso.uiuc.edu Illinois Mar 21 4:52 MY USUAL MIRROR
[will replace a file from ftp.riken.go.jp]
ftp.caldera.com Utah Mar 21 5:52 no change
[SURPRISE SINCE THIS IS NEWER]
ftp.cdrom.com California Mar 20 13:52 no change
ftp.kernel.org California Mar 20 22:52 no change
aij.st.hmc.edu California Mar 21 06:52 replaced
uiarchive.cso.uiuc.edu
ftp.fuller.edu California Mar 21 6:52 replaced
uiarchive.cso.uiuc.edu
>From this, one might conclude that a time greater than 1 hour results
in replacement.
I AM DISSAPPOINTED THAT WHEN I CHANGE DEBIAN MIRROR SITES,
THE PROGRAM "MIRROR" WILL OFTEN REGET THE WHOLE ARCHIVE.
QUESTION 2: Does "mirror" replace a file if the mirror site indicates
a file-date more than 1 hour newer than my computer's file-date,
irrespective of time zone?
QUESTION 3: When Debian went to version 1.2, many Debian mirror sites
mirrored ftp.debian.org badly, dropping good files and failing to
mirror others.
I expect this problem will recur in the future.
Is it reasonable to switch mirroring sites, or should I expect
such a switch to reget the whole Debian archive?
QUESITON 4: I suppose either my computer changed file-dates
or the archive site changed file-dates.
Which do you suppose happened, or do you suppose
system-dates/computer-clocks matter here?
Here are some relevant lines from my mirror.defaults file:
site=uiarchive.cso.uiuc.edu
remote_dir=/pub/systems/linux/distributions/debian
do_deletes=true
max_delete_dirs=50%
max_delete_files=50%
mode_copy=true
compress_conv_patt=
compress_conv_expr=
compress_patt=
# I do not set "force_times" (default is yes, forcing local times
# to match remote times).
# I do not set "get_newer" (default is true,
# so get archive files newer than my computer's files).
# I do not set "use_timelocal" (false would adjust to GMT, I gather).
--
Jim Burt, NJ9L, Fairfax, Virginia, USA
jameson@mnsinc.com http://www.mnsinc.com/jameson
"If merely 'feeling good' could decide, drunkenness would be the supremely
valid human experience." --William James, Varieties of Religious
Experience
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: