G'day all, Given that in Fedora and Linux in general, there are many different ways to accomplish the same task, I would like to ask people of their recommendations or experiences in what components to use for a small mail set-up for home and office and how they link together. From my reading so far an IMAP model would be good, but I am still unsure of the precise flow that mail would take in and out.
Currently I use what Fedora10 defaults to (sendmail and Evolution) for no reason other than they are the defaults, am not sure when one stops and the other starts and have not looked at anything else. I also use the Evo Spamassassin plugin and SA with remote filtering enabled which works well. We access 10 e-mail accounts all from the same ISP. Different people will need to read mail from their individual accounts, with some accounts being read jointly by 2-3 separate users. Volume of mail is low, eg. this list is 25-50% of total mail volume.
I have found much material on the individual components, but not on mail processing in general and how they fit together, complement, compete or overlap each other. Pointers to such process flow would be appreciated, as well as general opinions on piecing together such a project.
On Fri, 2009-03-20 at 10:12 +1100, Simon Slater wrote:
G'day all, Given that in Fedora and Linux in general, there are many different ways to accomplish the same task, I would like to ask people of their recommendations or experiences in what components to use for a small mail set-up for home and office and how they link together. From my reading so far an IMAP model would be good, but I am still unsure of the precise flow that mail would take in and out.
Currently I use what Fedora10 defaults to (sendmail and Evolution) for no reason other than they are the defaults, am not sure when one stops and the other starts and have not looked at anything else. I also use the Evo Spamassassin plugin and SA with remote filtering enabled which works well. We access 10 e-mail accounts all from the same ISP. Different people will need to read mail from their individual accounts, with some accounts being read jointly by 2-3 separate users. Volume of mail is low, eg. this list is 25-50% of total mail volume.
I have found much material on the individual components, but not on mail processing in general and how they fit together, complement, compete or overlap each other. Pointers to such process flow would be appreciated, as well as general opinions on piecing together such a project.
---- I use a setup that I implement on all my clients...
postfix MailScanner (wrapper for clamav & spamassassin) SQLGrey (greylisting) cyrus-imapd
mail is received by postfix, put into incoming queue, noticed by MailScanner, checked, passed back to postfix for delivery, and finally put into cyrus-imapd store.
cyrus-imapd includes server based sieve mail filtering, many automatic features and allows users or administrators to individually share mail folders. I set up all clients to use imap which allows them to use any program on any computer on the network and still access their e-mail and I only have to backup the server.
I use RHEL or CentOS for this mail server and not Fedora because of the need to frequently update. I use Fedora for user desktops.
Generally, when you want to run your own mail server, you also need to run your own DNS server because mail delivery depends upon DNS resolution and when you've got a mail server on your internal LAN, it has to resolve the domain name for e-mail delivery. Generally, I go for overkill on mail setup because e-mail is the office/workgroup essential. I also install the latest Horde/IMP/Kronolith/Nag/Turba/etc. (and LDAP) to give a fully integrated workgroup collaboration suite of not only shared e-mail but also shared contacts, shared calendars, etc.
If you plan on continuing to use the same mail accounts at the same ISP, then you would probably need to set up fetchmail (or getmail) to retrieve the e-mail from the ISP and feed it into your mail server.
Craig
Simon Slater wrote:
G'day all, Given that in Fedora and Linux in general, there are many different ways to accomplish the same task, I would like to ask people of their recommendations or experiences in what components to use for a small mail set-up for home and office and how they link together. From my reading so far an IMAP model would be good, but I am still unsure of the precise flow that mail would take in and out.
Currently I use what Fedora10 defaults to (sendmail and Evolution) for no reason other than they are the defaults, am not sure when one stops and the other starts and have not looked at anything else. I also use the Evo Spamassassin plugin and SA with remote filtering enabled which works well. We access 10 e-mail accounts all from the same ISP. Different people will need to read mail from their individual accounts, with some accounts being read jointly by 2-3 separate users. Volume of mail is low, eg. this list is 25-50% of total mail volume.
I have found much material on the individual components, but not on mail processing in general and how they fit together, complement, compete or overlap each other. Pointers to such process flow would be appreciated, as well as general opinions on piecing together such a project.
I have been running Scalix on Centos for a few years now. I am getting ready to bail out of that and am testing out SME (http://wiki.contribs.org/Main_Page). This is very stable and very functional for the SOHO, even SMB arena. I suspect there are several ways to have those 10 mail accounts picked up by the SME mail system and then presented to local users either via IMAP or POP3.
On Thu, 2009-03-19 at 16:43 -0700, Craig White wrote:
I use a setup that I implement on all my clients...
postfix MailScanner (wrapper for clamav & spamassassin) SQLGrey (greylisting) cyrus-imapd
mail is received by postfix, put into incoming queue, noticed by MailScanner, checked, passed back to postfix for delivery, and finally put into cyrus-imapd store.
The mail client (Evo, Kmail etc) then points to the local IMAP store?
cyrus-imapd includes server based sieve mail filtering, many automatic features and allows users or administrators to individually share mail folders. I set up all clients to use imap which allows them to use any program on any computer on the network and still access their e-mail and I only have to backup the server.
This is the type of functionality I'm after.
I use RHEL or CentOS for this mail server and not Fedora because of the need to frequently update. I use Fedora for user desktops.
We go at a slow pace with nothing on computer being "mission critical" so I'm happy to stay with Fedora for now while I'm still learning, but when I need to free up more time for other things CentOS is the favoured option and RHEL when I'm not hands-on much.
Generally, when you want to run your own mail server, you also need to run your own DNS server because mail delivery depends upon DNS resolution and when you've got a mail server on your internal LAN, it has to resolve the domain name for e-mail delivery. Generally, I go for overkill on mail setup because e-mail is the office/workgroup essential.
Sorting out the DNS now.
I also install the latest Horde/IMP/Kronolith/Nag/Turba/etc. (and LDAP) to give a fully integrated workgroup collaboration suite of not only shared e-mail but also shared contacts, shared calendars, etc.
Until now, only heard of the Evo suite and Koffice. Will check those out.
If you plan on continuing to use the same mail accounts at the same ISP, then you would probably need to set up fetchmail (or getmail) to retrieve the e-mail from the ISP and feed it into your mail server.
So the flow is: ISP -> fetchmail -> postfix/procmail/sendmail -> MailScanner -> SQLGrey -> Cyrus/Dovecot -> Evo/KOffice/etc
Craig
On Thu, 2009-03-19 at 20:19 -0400, Robert Moskowitz wrote:
I have found much material on the individual components, but
not on
mail processing in general and how they fit together, complement, compete or overlap each other. Pointers to such process flow would
be
appreciated, as well as general opinions on piecing together such a project.
I have been running Scalix on Centos for a few years now. I am getting ready to bail out of that and am testing out SME (http://wiki.contribs.org/Main_Page). This is very stable and very functional for the SOHO, even SMB arena. I suspect there are several ways to have those 10 mail accounts picked up by the SME mail system and then presented to local users either via IMAP or POP3.
SME looks interesting and looks like it would work well for another project in the pipeline that would use older hardware. For now I've put the effort into Fedora and want to continue in that vein. However where my knowledge has gaps would still fail me in SME methinks.
On Fri, 2009-03-20 at 12:00 +1100, Simon Slater wrote:
On Thu, 2009-03-19 at 16:43 -0700, Craig White wrote:
I use a setup that I implement on all my clients...
postfix MailScanner (wrapper for clamav & spamassassin) SQLGrey (greylisting) cyrus-imapd
mail is received by postfix, put into incoming queue, noticed by MailScanner, checked, passed back to postfix for delivery, and finally put into cyrus-imapd store.
The mail client (Evo, Kmail etc) then points to the local IMAP store?
---- yes, the one computer with all of the mail handling would be the IMAP server and all the computers would point to that computer. ----
I use RHEL or CentOS for this mail server and not Fedora because of the need to frequently update. I use Fedora for user desktops.
We go at a slow pace with nothing on computer being "mission critical" so I'm happy to stay with Fedora for now while I'm still learning, but when I need to free up more time for other things CentOS is the favoured option and RHEL when I'm not hands-on much.
---- suit yourself - RHEL is by design a long term supported platform. Current RHEL 5 seems to be spun from much of what was Fedora 6 ----
I also install the latest Horde/IMP/Kronolith/Nag/Turba/etc. (and LDAP) to give a fully integrated workgroup collaboration suite of not only shared e-mail but also shared contacts, shared calendars, etc.
Until now, only heard of the Evo suite and Koffice. Will check those out.
---- the idea is that the protocols are supported so if you have Kronolith set up, then you can point Evolution calendars on the server ----
If you plan on continuing to use the same mail accounts at the same ISP, then you would probably need to set up fetchmail (or getmail) to retrieve the e-mail from the ISP and feed it into your mail server.
So the flow is: ISP -> fetchmail -> postfix/procmail/sendmail -> MailScanner -> SQLGrey -> Cyrus/Dovecot -> Evo/KOffice/etc
---- Sendmail < or > Postfix procmail is pretty ugly and not needed if you use sieve Ingo (part of Horde) can give users a very nice sieve interface to control server based rules Greylisting not useful if you are doing fetchmail from ISP Cyrus IMAP < or > Dovecot
Craig
On Fri, 2009-03-20 at 12:12 +1100, Simon Slater wrote:
On Thu, 2009-03-19 at 20:19 -0400, Robert Moskowitz wrote:
I have found much material on the individual components, but
not on
mail processing in general and how they fit together, complement, compete or overlap each other. Pointers to such process flow would
be
appreciated, as well as general opinions on piecing together such a project.
I have been running Scalix on Centos for a few years now. I am getting ready to bail out of that and am testing out SME (http://wiki.contribs.org/Main_Page). This is very stable and very functional for the SOHO, even SMB arena. I suspect there are several ways to have those 10 mail accounts picked up by the SME mail system and then presented to local users either via IMAP or POP3.
SME looks interesting and looks like it would work well for another project in the pipeline that would use older hardware. For now I've put the effort into Fedora and want to continue in that vein. However where my knowledge has gaps would still fail me in SME methinks.
---- SME I believe is built from CentOS - again, it's all from packages that originated from Red Hat / Fedora so the setup is extremely similar.
SME tries to provide a turnkey setup and to gain ease of setup, you sacrifice features. It's a trade off only you can decide whether it's worth it or not. Myself, I would never do that.
Craig
On Thu, 2009-03-19 at 19:33 -0700, Craig White wrote:
If you plan on continuing to use the same mail accounts at the
same
ISP, then you would probably need to set up fetchmail (or getmail) to retrieve the e-mail from the ISP and feed it into your mail
server.
So the flow is: ISP -> fetchmail -> postfix/procmail/sendmail -> MailScanner ->
SQLGrey
-> Cyrus/Dovecot -> Evo/KOffice/etc
Sendmail < or > Postfix procmail is pretty ugly and not needed if you use sieve Ingo (part of Horde) can give users a very nice sieve interface to control server based rules Greylisting not useful if you are doing fetchmail from ISP Cyrus IMAP < or > Dovecot
Craig
Another question is: The ISP sends (or I collect) mail via POP3. Can this then feed into the local IMAP server? Or to phrase differently, is the mail, just mail to the various programs that handle it, irrespective of the transport protocol and so can pass through one protocol to another and from one program to another?
On Fri, 2009-03-20 at 14:19 +1100, Simon Slater wrote:
Another question is: The ISP sends (or I collect) mail via POP3. Can this then feed into the local IMAP server? Or to phrase differently, is the mail, just mail to the various programs that handle it, irrespective of the transport protocol and so can pass through one protocol to another and from one program to another?
---- fetchmail retrieves mail from ISP using POP3 or other supported protocols and then hands it off to your SMTP server (Sendmail or Postfix or ???)
Craig
Craig White wrote:
On Fri, 2009-03-20 at 14:19 +1100, Simon Slater wrote:
Another question is: The ISP sends (or I collect) mail via POP3. Can this then feed into the local IMAP server? Or to phrase differently, is the mail, just mail to the various programs that handle it, irrespective of the transport protocol and so can pass through one protocol to another and from one program to another?
fetchmail retrieves mail from ISP using POP3 or other supported protocols and then hands it off to your SMTP server (Sendmail or Postfix or ???)
It can also do a rewrite of the email address as part of the process, so your local email address does not have to match the ISPs email address. It can also handle cases where your ISP puts all the mail for the users on your local domain into one mailbox.
Mikkel
On Fri, 2009-03-20 at 09:18 -0500, Mikkel L. Ellertson wrote:
Craig White wrote:
On Fri, 2009-03-20 at 14:19 +1100, Simon Slater wrote:
Another question is: The ISP sends (or I collect) mail via POP3. Can this then feed into the local IMAP server? Or to phrase differently, is the mail, just mail to the various programs that handle it, irrespective of the transport protocol and so can pass through one protocol to another and from one program to another?
fetchmail retrieves mail from ISP using POP3 or other supported protocols and then hands it off to your SMTP server (Sendmail or Postfix or ???)
It can also do a rewrite of the email address as part of the process, so your local email address does not have to match the ISPs email address. It can also handle cases where your ISP puts all the mail for the users on your local domain into one mailbox.
Mikkel
I'm still coming to terms with the terminology, is one mailbox equivalent to one e-mail account with ISP or one directory where all e-mail from all sources is put?
The point on rewriting headers brings up another question on address names and mailer apps currently running. Having a LAN with private ip range, the local boxes can have FQDNs like dell.local or acer.local and the world at large doesn't care. At the moment sendmail and Evo are running (just whatever the install defaults to, not on my preference or choice), so if I use mutt to read root's maillogs, is that using sendmail at all or is mutt just looking into the directory containing the mail? Does Evo use sendmail, or does it fetch from the ISP directly? Can Evo read the local mail (simon@dell.local)?