In article <1992Oct1.202223.22993@netcom.com> wisner@netcom.com (Bill Wisner) writes:
>barmar@think.com (Barry Margolin) writes:
>>The context of the suggestion was that everyone is NFS-mounting
>>/usr/spool/mail from the mail host.
>
>Here, everyone mounts /usr/spool/mail from a server that *isn't* the
>mail host -- each workstation handles its own mail.
Then the recommendation to us the OR option may not apply to your
situation. The original post said, "if you're mounting /usr/spool/mail
from the mail host, use the ORmailhost option," and someone (you?) then
said that this may not be a good idea because of its effecs on the From:
line. Given the precondition in the recommendation, that effect seems
desirable, not a problem.
>>In that case, why wouldn't it be
>>correct for the From: address to be an address that gets delivered to that
>>host?
>
>Because the server from which /usr/spool/mail gets mounted doesn't
>necessarily know about every user on every workstation.
The original posting was recommending against delivering (as opposed to
reading) mail this way. The locking mechanisms used by mail delivery
software may not be very reliable over NFS. Therefore, if two machines are
delivering mail to the same user simultaneously there may be problems.
>NIS, you say? Yes, we use an NIS passwd map -- but our smallest NIS
>domain has nine servers, and any given user's mail spool is contained
>on precisely one of those servers.
How does a machine know which server to mount /usr/spool/mail from? Is
each machine customized to a particular user? You probably have an order
of magnitude more workstations than we do, so I don't know how you manage
extensive per-machine customization like that. We have several hundred
machines, and can only manage the complexity by having standard
configurations for each class of machine (user workstations, compute
servers, and file/NIS servers). Any per-user customization is in personal
files (e.g. .login, .cshrc), not machine customizations.
I'm not sure what we'll do if we ever get large enough that /usr/spool/mail
can't fit in a file system. I guess we could have departmental
/usr/spool/mail file systems, so mail would be addressed to
user@<dept>.think.com. A central aliases file on the think.com would allow
mail to user@think.com to be routed without senders having to specify the
department.
>We could start maintaining lists of which users are on which servers
>and have servers handle mail rather than individual workstations, but
>that would simply add another layer of complexity and difficulty to
>our jobs, and our present system works quite well.
I notice your address is wisner@EBay.Sun.COM. Is EBay the NIS domain or
the actual server? If it's the domain (which seems more likely to me), how
is it then directed to the appropriate mail server without such a list?
Or are you talking about the setup at netcom.com? Again, you didn't
specify a specific file server in your address, so my question stands.
-- Barry Margolin System Manager, Thinking Machines Corp.barmar@think.com {uunet,harvard}!think!barmar
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:50 CDT