Re: I've thought better of Linux
- From: Bulent Murtezaoglu <bm@xxxxxxx>
- Date: Sat, 18 Jun 2005 19:15:51 +0300
>>>>> "UH" == Ulrich Hobelmann <u.hobelmann@xxxxxx> writes:
[...]
UH> After all, what huge amounts of work does a mail server do?
I wouln't say huge, but it does more than you imply.
UH> Not THAT much, accepting SMTPs,
Except sendmail was supposed to handle other protocols and do the right
thing including things like accepting UUCP mail, and relaying it
through SMTP gateway to a bitnet host etc. While I am not fond of the
odd sendmail.cf language (which M4 does hide, but M4 itself is no monument
to elegance either) there is a reason why it is there.
UH> putting it into files for local users,
Except it when it doesn't. The mail can be properly local in the
lowest visible MX sense with the spool being elsewhere, or the
localness might be decided by some failure-prone mechanism (LDAP, split
DNS MX records, NIS etc.). Or maybe there is no spool for the user
indicated? (eg an alias, autoresponder with its own rejection
mechanisms, or a list etc.)
UH> ... having an outgoing queue that gets delivered if
UH> the remote server is responsive
Except you can't do this in a straightforward manner. You have to
refrain from creating a thundering herd problem, deal with flaky or
misconfigured DNS (so perhaps you know the name of the machine you'd
like to send to, but you cannot get the IP. What now? AFAIR Qmail
would fail in odd ways in this sceario and would bounce deliverable
mail.) etc. etc.
UH> then some firewall-like
UH> filtering options, like what mail to reject.
Which of course is a non-trivial problem if you want to do it right.
And remember my 'right' is different than your 'right.' The customer
is always right but it doesn't mean they know what they want or
communicate it properly even when they do. And there is more than one
of them.
UH> Where's the
UH> immense inherent complexity of the application domain? Where
UH> would it be in a web server, or in an FTP?
I don't know if I'd call it _immense_ but managing to get mail to be
realiable in a hostile network where your peers might suffer from
all manner of brokenness is more complex than ftp or http.
[...]
UH> But for all non-exotic setups I expect to be able to config a
UH> server for *anything* in an hour. [...]
I don't expect any such thing unless I have had some success or,
preferably some well-understood _failures_ in doing so previously.
My fear would be getting something working _by accident_ and having to
learn what that accident was when the thing fails at a critical time.
I am _not_ saying there isn't any gratuitous complexity in sendmail, or
that the present mail architecture is well-designed. But for the world
we are living in such things don't 'just work' after being configured in
a DWIM fashion. Otherwise reasonable expectations don't change that
fact.
cheers,
BM
.
- References:
- I've thought better of Linux
- From: Brandon J. Van Every
- Re: I've thought better of Linux
- From: Ulrich Hobelmann
- Re: I've thought better of Linux
- From: David Magda
- Re: I've thought better of Linux
- From: Matthias Buelow
- Re: I've thought better of Linux
- From: Ulrich Hobelmann
- Re: I've thought better of Linux
- From: Matthias Buelow
- Re: I've thought better of Linux
- From: Ulrich Hobelmann
- Re: I've thought better of Linux
- From: Matthias Buelow
- Re: I've thought better of Linux
- From: Tim X
- Re: I've thought better of Linux
- From: Ulrich Hobelmann
- Re: I've thought better of Linux
- From: Tim X
- Re: I've thought better of Linux
- From: Ulrich Hobelmann
- I've thought better of Linux
- Prev by Date: Re: I've thought better of Linux
- Next by Date: Re: I've thought better of Linux
- Previous by thread: Re: I've thought better of Linux
- Next by thread: Re: I've thought better of Linux
- Index(es):
Relevant Pages
|
Loading