brian@sdcc3.UUCP (Brian Kantor) (10/16/84)
A few months ago I inquired of the net what sort of local addressing
could be used to isolate our local network names from the "flat
namespace" of the Internet and uucp domains. We are slowly (in my
spare time, of which I have little) working on something that will do
the following.
when I send mail from one of our campus machines ('wizard', for
example), the address will appear as follows:
sdcsvax!brian%wizard
since 'sdcsvax' is our portal to the world.
UCSD is our local domain. (We have a ucsd net registered with the NIC).
When we get full sub-subdomains installed (there are about 50 machines
on campus that can send/receive mail), split along administrative or
departmental lines, each sub-subdomain(ssd) will have a name, and a mail
server that can route mail within that ssd. So, for example, the
machines in the computer center will generate addresses of the sort
brian%sdcc9.cc.ucsd
as a return address. Mail entering our portal at sdcsvax can be
addressed using this address, eg
sdcsvax!brian%sdcc9.cc.ucsd
or it can just omit the name of the computer system and use just the
domain, ie
sdcsvax!brian%cc
which sdcsvax will be able to recognize and will route to the domain
server for the computer center, which will in turn be able to discern
the proper machine and mailbox for final delivery.
The great advantages of this scheme is that we will be able
1. to add machines, or change their names at will, without worries of
name conflicts with other sites.
2. allow each of our internal administrators handle their own userid
assignments, without having to consider campuswide name databases.
3. greatly simplify the addition of a name server (thus allowing us
to use addresses such as "brian kantor"%cc, which is pretty easy to
remember if we use the generally accepted abbreviations for the
departments (abbreviations that most people use already).
The disadvantages are
1. The sendmail configuration files are going to be a nightmare.
2. If a server goes awry (down, or gets confused) an awful lot of
mail is going to get lost.
3. We're awfully dependent upon other sites not munging addresses
with % and . in them. RFC822 says they won't, but this is the real
world.
If/when we get on the Internet, the above composite address
(user%machine or user%ssd) will simply be the mailbox part of a
standard RFC822 address, such as brian%wizard.cglab@UCSD, or more
simply, brian%cglab@UCSD.
Can anyone see any problems with this scheme? Am I totally wrong?
Is it going to work if I can figure out how to do it? Please send me
mail with your ideas/comments/flames.
ihnp4 \ Brian Kantor, UC San Diego
decvax \
akgua >---- sdcsvax ----- brian
dcdwest/
ucbvax/ Kantor@Nosc