On Saturday 05 November 2011 07:50:58 /dev/rob0 wrote:
> I have cut all the irrelevant and whiny crap from the quotes, and I
> ask that others please not continue that off-topic and useless
> discussion. One part of this, q.v., deserves to be addressed.
>
> On Saturday 05 November 2011 09:03:18 David Southwell wrote:
> > On Saturday 05 November 2011 06:42:12 Simon Brereton wrote:
> > > On 5 November 2011 08:21, David Southwell
>
> > > <ad...@vizion2000.net> wrote:
> snip
>
> > > > Clearly postfix is need of an intelligent parser that will to
> > > > pinpoint errors such as this in master.cf and main.cf. That is
> > > > because stupid computers are better at parsing chores than
> > > > human beings.
> > >
> > > Postfix has such a parser - which is why the documentation points
> > > out that lines should not start with a white-space.
>
> snip
>
> > You are failing to distinguish between a diagnostic parser and an
> > executive parser. An executive parser rejects incorrectly
> > configured lines at runtime. A diagnostic parser would tell you
> > that there is an excess space at a specific location. A really
> > good executive parser would also log the location of incorrectly
> > configured lines to facilitate the work of an administrator.
>
> And that would be far more difficult than you imagine.
How do you know how much I imagine. What makes you believe that I do not know
it is difficult!
The problem you identify in subsequent lines, has its roots in postfix's
rather primitive formatting structure.
If it were replace by something like:
{submission (variant,modifier [connector] data )
(variant = data)
(variant = data)
end submission
}
This type of formatting structure (it would need a few more symbols to cover
all the current alternatives) is easier for humans to read, makes clear the
separation between modules and facilitates the building of diagnostic &
executive parsers to test, implement and log outcomes.
IMHO the problem caused by retaining the earliest forms of formatting known to
unix is what presents postfix users unnecessary challenges.
It is easier to change the formatting structure than map a parser to the
current idiosyncratic framework. It would not be necessary to reinvent
Postfix's executive parser because it would not be that difficult to build a
diagnostic parser which could also convert a new format into the existing
format.
Idiosyncratic formatting is a curse inflicted on system administrators who are
expected by those who are dedicated to supporting a single application. The
demands they make on administrators are therefore unrealistic.
