On Wednesday 28 September 2005 01:10, Steve Lamb wrote:
> As I said, and what do those programs do when said helper program
> isn't available? By definition they have to have a queue of their own
> since the only thing they can control is their own behavior. To not
> have some method of dealing with an external failure is bad design,
> plain and simple. And since any good design is going to be resistant
> to external failure it doesn't matter through what means that external
> failure is reached.
By building an MTA's functions into every program that needs to send mail?
> I guess a better question to ask is why everyone insisnts on having
> an external queue when an internal one is pretty much required for
> robustness?
I don't understand why you say that. Many programs on Unix-like systems
rely on the functionality and availability of other services. It's part
of the tradition and philosophy (again, see ESR's Art of Unix
Programming).
I think a better question is: why should cron, at, mailx and all the MUAs
independently be coded and debugged to implement the functionality that
any one MTA can provide?
