I'll summarize the process for you here to get you started.
There are three parts to a basic mail server: - Mail Transport Agent (MTA) - Local Delivery Agent (LDA) - Mail User Agent (MUA)
In other words, three separate programs actually work in series to comprise a mail server. Each one performs a specific function, they are all interchangeable, and you must have all three for the mail server to function. Here is what they all do, briefly:
The MTA (usually sendmail) transfers mail between servers. The sending machine's MTA looks at the To: header of the message, opens up a connection to the MTA on the receiving domain's mail server, and sends the message. The receiving machine listens on the configured port(s) for mail connections from servers in the outside world. Usually SMTP (Simple Mail Transfer Protocol) is used by the MTAs to communicate the transaction.
On the receiving end, the MTA passes the message off to the LDA, which figures out which user on the local machine to deliver to, and puts it in their mailbox. Here is where extra filtering utilities such as spamassassin and virus blockers are sometimes tacked on to make sure only valid mail actually gets delivered.
In this basic model, the recipient then opens up their MUA (pine, mutt, etc.) on the receiving machine, which looks in the user's mailbox, sees the message, and opens it for viewing. To send mail, the user composes a message using the MUA. When the user tells the MUA to send the message, it passes the new message directly to the machine's MTA, which opens a connection to the receiving domain's mail server and sends the message.
This basic model was common in the early days of the internet, before individual workstations: users would work directly on the server via a dumb terminal, so the mail didn't need to go any further than the receiving server.
Nowadays, however, most people don't use MUAs directly on the server -- they check their mail from their workstation. I don't go down to a server farm in Cupertino to check my email directly on the server. :-) So to extend the MUA reach, we now use remote-mail protocols like POP and IMAP, which provide an interface for MUAs on client machines (i.e. non-mail-serving machines) to remotely interact with a mail server's mailbox. They are set up in addition to the three core components, and are simply a means to allow a mailbox to be checked from a MUA on another computer. The remote-mail protocol isn't technically one of the core mail serving tools, because you can get mail perfectly well without one, but because nowadays most people use a workstation separate from their mail server, it's essentially an integral part of the mail server.
I think Fetchmail was also mentioned in the thread. Fetchmail is a special tool which combines the functionality of a remote-mail protocol and an MTA. You can use it to regularly download mail from multiple mailboxes on multiple servers (probably using POP) and pass them off to the LDA on your local server. So if you have a bunch of email accounts -- for instance, a school account, hotmail account, .Mac account, and work account -- you can use Fetchmail to periodically retrieve new mail from all four accounts and deliver them to accounts on your own mail server. Then you can use an MUA to read mail from all your accounts on just one server.
So a nicely functional mail server would consist of the following:
- sendmail for MTA (activates SMTP service on port 25)
- fetchmail to retrieve mail from all your accounts (uses a remote-mail protocol for retrieval from mailboxes on other servers)
- procmail for LDA (delivers messages to wherever you've configured the local server to keep mailboxes, in the format you've specified)
- spamassassin to filter the junk
- virus detection filter for protection
- IMAP service for remote mail access (activate using xinetd, uses port 143?)
- your favorite IMAP-compatible MUA on your workstation
The domain hosting part adds a little complexity. There's probably a way to host a domain on a static IP address, but I don't know how to do it, and I'd imagine it would be rather complicated and unreliable. So I would go with a static IP address if at all possible.
Assuming you do get domain hosting up and running, you'll then be able to receive email directly to your domain (e.g. [EMAIL PROTECTED]). If not, you'll only be able to retrieve mail from other accounts using Fetchmail, not receive email addressed specifically to your machine. Of course, as long as you're running sendmail and it's configured to receive incoming mail, you'll always be able to get [EMAIL PROTECTED] (your IP address) -- but that's of no use to anyone else when you have a dynamic IP address, because your email address would constantly be changing.
As for the difference between the remote-mail protocols: with POP, message storage on the server is transient. As soon as the client MUA downloads new messages, they are removed from the server's mailbox, and stored only on the client's machine (i.e. within Outlook, Entourage, Evolution, KMail, etc.). IMAP, on the other hand, stores all messages on the server, and provides an interface to the client MUA to view messages kept within the server mailbox. Most MUAs download all the mail from the server anyhow to cache it, but the server retains all messages until they are explicitly deleted.
Hope that helps. If I can answer any other questions, don't hesitate to ask!
Cheers,
Morgan
-- redhat-list mailing list unsubscribe mailto:[EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-list
