[comp.unix.ultrix] YP

doug@tci.bell-atl.com (Doug Mildram) (11/15/89)

To the 3 who replied, thanks for the tips on yellow pages.
No one liked it.  They use simple copy techniques for nitely distribution
of /etc/passwd , (/etc/hosts, whatever...).  Just make sure folks don't 
change their passwords on machines whose /etc/passwd's get rewritten nitely.

YP does seem overly complex.  Anyone care to defend it?
-- 
Doug Mildram 	  Technology Concepts Inc.
40 Tall Pine Dr.  uucp:	{anywhere}!uunet!tci!doug, doug@tci.bell-atl.com
Sudbury MA 01776  (508) 443-7311 x273     

tom@osf.osf.org (Tom Jordahl) (11/16/89)

In article <431@tci.bell-atl.com> doug@tci.bell-atl.com (Doug Mildram) writes:
>To the 3 who replied, thanks for the tips on yellow pages.
>No one liked it.  They use simple copy techniques for nitely distribution
>of /etc/passwd , (/etc/hosts, whatever...).  Just make sure folks don't 
>change their passwords on machines whose /etc/passwd's get rewritten nitely.
>
>YP does seem overly complex.  Anyone care to defend it?

Well,
YP is really nice *when it works*.  It gives you the ability to
maintain a central database at a central machine, without having
a central point of failure for all your (1-500) workstations.

The problem with using a "simple" copy technique to maintain a distributed
database are numerous.  What if joe-schmo's workstation is turned off
or down when the copy scripts run?  He has to wait till the next run.
If you run the distribution often, there's overhead involved and passwd
et al doesn't change *that* often.  You can keep track of who you have
sucessfully sent the new files too, but at this point it starts becomming
less "simple".

Yellow pages also allows the individual workstation "owner" to maintain
local information in their /etc files, guest accounts, personal host
addresses, etc. (no pun intended)  This is in addition to the global
info.  This info can also override the global information, allowing
a user to have a local home directory on his (diskfull) station, but
having the same environment on all the other machines.

On a large network, you can have several servers, any of which can serve
any client.  The databases are guaranteed to be updated, although there
is some delay.  It is also possible for the master to push the database
out, or the clients to pull the database in whenever the user finds
the update too slow.  The updates only happen if the databases are
modified since the last update, so unneccesary trafic is avoided.

On the down side:  If YP is being grumpy, your clients can get hung
in a bad way looking for a server they insist is still around.
Propagation of the databases in some implimentations is not
reliable. (This is *not* true for Suns, in both 3.5 and 4.0 I have
seen it work flawlessly.)

Also on the down side, its a pain.  You can have a separate source
directory on the master host, which everyone may not know where to
find.  Having text files of the /etc information was designed to
be convient, and it is.  YP puts the info into dbm format, and you
must use ypcat and the like to get at the info.


Bottom line:  When it works, it works well.
              When it doesn't work, you have a big mess.

>Doug Mildram 	  Technology Concepts Inc.


Tom Jordahl                                       Internet: tom@osf.org
Open Software Foundation                          UUCP:   ..!uunet!osf!tom
       "A silly quote is only worth what you put into it."  -- me