Questions and Answers
Bjorn Satdeva
The InterNIC, which is the office that keeps track of
assigned IP
addresses and domain names, apparently continues to
be in a mess. On one
hand, they have their work cut out for them with the
enormous growth in
the number of second-level domain names over the past
2 years. On the
other hand, they started to charge a yearly fee for
this service a year
ago. I do not want to restart the discussion of the
appropriateness of
this, especially as a modest fee for this service probably
is
appropriate. However, the income from those charges
was supposed to be
used to improve the service from the InterNIC and also
be invested into
backbone. Lately, I have had a small problem with the
InterNIC, as I got
charged for a fee for sysadmin.com, dating back to 1989,
the first year
of that domain's existence. When I inquired about this
and asked if they
now had a policy of charging domain name holders retroactively
to the
date of creation of their domain, I never got an answer.
I did however
get a new invoice (still by email, in spite of my request
to receive
this by U.S. mail), this time for the current year,
although it is dated
back in March.
The InterNIC has a large task in front of them and is
working on
inproving their system to provide a more reliable service.
However, it
may serve the rest of us well to stay on top of how
they handle our
domains. I know of at least one case in which a domain
was changed over
to a new organization in spite of the previous organization's
having had
it for several years. So for now, look out.
I have forwarded to MIC the required paperwork for
a change in our
domain registration, but don't recall any place where
they ask for MX
hosts that will point to this domain. Our site receives
email through
UUCP, and we need this information in our domain configuration.
Did I
miss something?
MX records are not handled by the InterNIC. A UUCP
site can send and
receive email from sites that are on the Internet using
what is called a
forwarder. This was a common practice a few years ago,
when direct
Internet connections were less common and much more
expensive than they
are today. UUCP is still around, though, and will probably
be for some
time, as it is inexpensive and all the needed software
comes with the
system. Let me, therefore, briefly explain the steps:
You will need a forwarder, which is a site that is directly
connected to
the Internet, and that is willing to intercept email
for your site and
forward it to UUCP. To do this, they need to make a
slight change to
their sendmail configuration, so any email they receive
for your site
will be queued up for transmisson by UUCP.
The rest of the Internet will know to route email for
you to your
forwarder through the MX record for your domain. MX
records (Mail
eXchange) are nameserver records used for directing
the flow of email.
These records have other purposes than just handling
UUCP forwarders,
but the general mechanism is the same. The MX record
for your domain
will point to your forwarder, and the mail delivery
agents on the
Internet will therefore know to route email for you
to your forwarder.
This is completely transparent, and the users who send
the email will be
unaware of the fact that they are sending email to a
UUCP site.
The Internic is not directly involved in the MX record
business. They
only provide the information that goes to the root servers,
and that
information only contains a pointer to nameservers that
know about your
domain. Those nameservers will need to have an MX record
pointing to
your forwarder. Typically, the forwarder will be the
maintainer of the
nameserver that contains your MX record, just because
this is a
practical way of doing it. Technically, your MX record
can be handled by
any nameserver (as long as the root nameservers have
delegated your
domain to them).
This scheme will allow anybody on the Internet to send
email to you by
actually sending to your forwarder, using the MX record.
This is all
done automatically. The only problem is that your email
addresses will
still use the ugly old UUCP bang (!) addresses. If you
get smail3 and
use it instead of sendmail, it can use what is known
as "batched SMTP"
for sending and receiving email. If your forwarder supports
batched
SMTP, you will be able to send and receive email, using
SMTP end to end,
and nobody will be any wiser.
I have been told that my nameserver contains "Lame
Delegations." What
is that?
A "lame delegation" is when you delegate
authority for a domain or part
of a domain to a nameserver which does not have that
information.
Delegation of authority is what makes the nameserver
paradigm work. For
example, the root nameservers have delegated the domain
sysadmin.com to
my nameserver heimdal.sysadmin.com.
The typical lame delegation occurs when nameserver N
is delegated as
authoritative for domain D. However, when you ask N
about D, it returns
nonauthoritative data or no data at all. Nonauthoritative
data is
nameserver speak for, "I have the information you
asked for, but it is
second-hand information."
The real problem with lame delegations is that they
can generate
substantial traffic on the Internet, but without being
able to resolve
the names or IP addresses that the users of the name
service are
attempting to get.
The Domain Name System FAQ offers a more technical examination
of "Lame
Delegation." It is available at the system administration
ftp archives
at:
ftp://ftp.sysadmin.com/pub/dns
I read the man pages and my O'Reilly book on the dump
command, and just
can't get it right. All I want to do is dump a directory
to a file on
another disk. I am testing the capacity of a 9 Gb disk
with the dump
command. I already used the tar command. I want to copy
the /usr
directory to another mounted disk. It seems simple,
could you help or
point me to a place that will explain?
What you are trying to do is to use dump to archive
some files.
However, dump was not written as an archiving program,
and does not work
well when used for that purpose. Use dump for your regular
backups, and
tar for archiving purposes. It does not matter for this
discussion if
your image is on tape or on the hard disk.
If you are a bit confused about the difference between
backups and
archives, you are not alone. Your backup is a (hopefully
daily) snapshot
of your computer. It does not care about the purposes
of the data, as
the goal is to be able to do a full or partial restore
of the data
contained on the disk in case of a disaster of some
kind. Note that
disaster has a wide meaning here; it can be a disk crash,
a user
accidentally removing important files, or it can be
a cracker that
intentionally alters data. The purpose of the backup
is to make sure if
you lose data you can get it back as fast as possible.
Archives, on the
other hand, are typically snapshots of data related
to a single or group
of users. One example is an archive containing all the
files related to
the development of an application. You can restore one
or more lost
files from an archive tape, but the main purpose is
to get a snapshot of
the state of the application on a given date (i.e.,
when a certain
milestone has been reached).
When copying directories, you can use tar, but if the
directory is
identical to a filesystem, you are again better off
using dump. For one
thing, most versions of tar are unable to handle files
with "holes,"
such as dbm files. When restored, the file's real size
will become
identical to its virtual size, which often is many,
many times bigger,
making it impossible to restore such files to the hard
disk.
To use dump for copying an entire filesystem, you can
use
dump 0f - /usr | (cd /mnt; restore xf -)
This will cause dump to write the content of the dump
archive to stdout,
and read by restore from stdin. You can create a similar
pipe from tar,
but make sure that you set the correct options for tar,
so owner, group,
access mode, and last modification time is preserved
(the specific
options vary slightly between versions, so check your
man page).
About the author
Bjorn Satdeva is the president of /sys/admin, inc.,
a consulting firm
which specializes in large installation system administration.
Bjorn is
also co-founder and former president of Bay-LISA, a
San Francisco Bay
Area user's group for system administrators of large
sites. Bjorn can be
contacted at /sys/admin, inc., 2787 Moorpark Ave., San
Jose, CA 95128;
electronically at bjorn@sysadmin.com; or by phone at
(408) 241-3111.
|