I'm not sure what a link error would be in a language like sieve;
was thinking that all implementations would be heavily based on a
lisp-like engine of some sort, but perhaps I misunderstand it. I
agree that syntax errors should not be exceptions. Further, I think
syntax errors should be caught before interpretation begins.
My personal view is that syntax errors should be caught before the
script is put to work. I am not willing to equal this particular
situation with avarage execution of command shell scripts.
These scripts are about to execute in batch mode as part of a
delivery process.
An good example is aliases as used by sendmail. The aliases file
have to be processed before being put into the message dispatch
process. Syntax errors are reported to the operator. The alias
lookup mechanism in sendmail does not have to deal with
syntactical errors.
Why not apply a similar requirement on filtering?
Hmm, perhaps I wasn't clear; I'm not advocating the adoption of a
Python-like syntax for sieve. I was just trying to clarify the point
about Python syntax. (Which is not to say that I think
indentation-based syntax is bad; I like it. It's just not that
important for something like sieve.)
I tend to agree, with one exception(!):
I thought of Sieve as potentially being exposed to the end user.
In that case, indentation based syntax can make a difference.
Otherwise I agree, it's not that important.
--
Tomas Fasth