[buug] Miscellaneous stuff (2005-09-01 BUUG meeting, etc.)

[Michael Paoli]

Some random bits from the 2005-09-01 BUUG meeting, and related stuff:

Some stuff we discussed: LINUX and live CDs
For those not as acquainted with installing LINUX, and interested in
installing it or checking it out a bit,
Ubuntu Linux http://www.ubuntulinux.org/
and
Knoppix http://www.knopper.net/knoppix/index-en.html
may be good places to start.
With a "live CD", one can boot and run it form CD-ROM, without installing
it.  This is particularly handy for checking it out, seeing how well it
supports one's hardware, and without need to alter one's hard drive contents.
Ubuntu has both a "live CD" and an "install CD", so one can check out the
live CD, and if things seem to work well and one likes it, one can then use
the "install CD" to install it onto the hard drive - with same hardware
detection and support, etc., so such an install should work quite similar
to what one can effectively test and preview on the "live CD".
Knoppix is much more "live CD" oriented, though it does offer a hard drive
installation option (script/program) on the CD.
Both Ubuntu and Knoppix are
Debian http://www.debian.org/
based distributions.
Debian is very well supported, and generally quite highly recommended (it's
also certainly my favorite) but it may not be the easiest starting
point.

Backups - various things discussed, including:
tar(1)
cpio(1)
pax(1)
dd(1)
bacula http://www.bacula.org/
amanda http://www.amanda.org/
Is there an open source solution for hot backups of NTFS filesystems
that includes not only all the data, but all the metadata (e.g. all the
NTFS ACL and related information)?  If not, are there sufficient open
source solutions/components to feasibly construct such a solution (e.g.
are there Perl modules that know how to read and write NTFS ACL information)?
I haven't looked at open source live/hot NTFS backups much, but quick check
via google does find items such as:
http://lists.samba.org/archive/samba/2002-April/041706.html
On the matter of backups, we mentioned some things we don't like.
Here's a little demonstration of something I mentioned I don't like about
tar(1).  Most notably, one obtains a rather non-intuitive result - if a
file has multiple hard links, and one subsequently restores a non-first
occurrence of the link from the tar archive, the action of tar is to
create the link, not to restore file data.  Example:
$ ls
$ echo old > a
$ ln a b
$ ls -ion a b
110128 -rw-------  2 1003 4 Sep  2 20:51 a
110128 -rw-------  2 1003 4 Sep  2 20:51 b
$ cat b
old
$ tar -cf tar a b
$ rm b
$ ls
a  tar
$ echo new > a
$ cat a
new
$ tar -xf tar b
$ ls -ion a b
110128 -rw-------  2 1003 4 Sep  2 20:52 a
110128 -rw-------  2 1003 4 Sep  2 20:52 b
$ cat b
new
cpio(1) doesn't have this particular "issue":
$ rm a b tar
$ echo old > a
$ ln a b
$ cat b
old
$ ls -d [ab] | cpio -o -H newc > cpio
1 block
$ rm b
$ echo new > a
$ cpio -im -H newc b < cpio
1 block
$ cat b
old
$ cat a
new
$ 
Essentially, with cpio(1), when multiple hard links on a file are backed up in
an archive, all those that are restored at the same time will be linked and
will contain the data that was backed up, but will not be linked to pathnames
that are not extracted but that were the same inode (an additional hard link)
when the cpio archive was created.  Perhaps just my opinion :-), but I consider
the cpio(1) behavior to be the more expected and appropriate behavior
(satisfies principle of "least surprise", etc.).
Also mentioned, GNU tar can do some semi-intelligent handling of sparse files.
It would seem the only common (as in supplied with the operating system)
means of getting sparse files precisely right, is by dd or one of the
dump(8) family of commands.  GNU tar(1) Example:
$ rm *
$ dd if=/dev/null of=allsparse bs=1024 seek=1024
0+0 records in
0+0 records out
0 bytes transferred in 0.000044 seconds (0 bytes/sec)
$ dd if=/dev/zero of=notsparse bs=1024 count=1024
1024+0 records in
1024+0 records out
1048576 bytes transferred in 0.020337 seconds (51560078 bytes/sec)
$ dd if=/dev/null of=semisparse bs=1024 seek=512; dd if=/dev/zero bs=1024 count=
512 >>semisparse
0+0 records in
0+0 records out
0 bytes transferred in 0.000894 seconds (0 bytes/sec)
512+0 records in
512+0 records out
524288 bytes transferred in 0.010161 seconds (51598060 bytes/sec)
$ ls -ons
total 1536
   0 -rw-------  1 1003 1048576 Sep  2 21:26 allsparse
1024 -rw-------  1 1003 1048576 Sep  2 21:26 notsparse
 512 -rw-------  1 1003 1048576 Sep  2 21:28 semisparse
$ tar --sparse -cf tar *sparse
$ rm *sparse
$ tar --sparse -xf tar
$ ls -ons *sparse
   0 -rw-------  1 1003 1048576 Sep  2 21:26 allsparse
1024 -rw-------  1 1003 1048576 Sep  2 21:26 notsparse
   0 -rw-------  1 1003 1048576 Sep  2 21:28 semisparse

We also discussed some "intellectual property" political, law
(copyright, trade secret, patent) and cultural stuff, "public domain",
etc., and even including some rather early history of copyright law in
Britain.

Question was asked about jobs in open source.  I mentioned there may be some
stuff in various User Groups' lists/archives (e.g.
http://lists.balug.org/pipermail/balug-talk-balug.org/
).  I know for example, I think when relatively recently (a few months back)
a presentation was done by xensource (www.xensource.com), and if I recall
correctly, they had opening(s) (may have been primarily for kernel hackers).
Some User Groups also have lists specifically related to jobs/employment.
Starting here:
http://linuxmafia.com/bale/
and drilling down into the various sites and their archives/lists might
be useful.  Note also that sometimes stuff will be mentioned at a LUG meeting,
and not make it to the LUG's list(s), so making at least some meetings, and
expanding contacts can be useful.  Lists and/or meetings can be useful
for announcing opening(s) and/or availability (current or upcoming).  Note
that various User Groups may had different policies/procedures on what
should/shouldn't be posted/announced when/where/how, etc.
Stuff like Craig's List http://www.craigslist.org/
and various professional associations/groups, e.g.:
Usenix http://www.usenix.org/
SAGE http://www.sage.org/
http://www.icca-sfba.org/
may also be useful.