[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: snmpconf language implementation




I'd like to echo David using different words:

We may think these things are defects (or serious defects), but I am
impressed by an important fact: The C and C++ specs have been updated
several times over the years by a group of esteemed experts in that
field and each time they declined to change these aspects of the
language. Maybe they know the fix is worse than the disease. Maybe they
know that these really aren't defects at all. We are esteemed experts in
a *different* field. I'm loath to substitute our judgement for theirs in
this matter.


Steve


David Partain wrote:
> 
> Greetings all,
> 
> Ray Byler wrote:
> rb> I still think nested comments should be allowed.  We are not
> rb> re-implementing C, we are creating a new language, and that
> rb> gives us the chance to remedy defects in C.
> 
> As Joel said, no it doesn't.  While we may consider this a
> blatantly obvious and trivial thing, such things are to some
> degree in the eye of the beholder.  This is a dangerous, very
> slippery slope that we've been (very) rigorous about avoiding.
> While this may appear stubborn and stupid, there have been
> numerous "wishes" for "improvements" to the languages we are
> stealing from that have previously been removed from the table,
> because the rule is: we are "not inventing a new language."
> 
> I'm sure that if we were to begin down this path, we'd get
> numerous things that the community considers bugs in C/C++
> and we'd finish the language in 2004. :-)
> 
> The primary reason for this as I see it is that we are
> (network) management people, not language people (I know there
> are exceptions).  We should defer to language people in the
> design of PolicyScript.
> 
> rb> Not allowing nested
> rb> comments is a defect.  Many C compiler implementors recognized
> rb> this, and created their compilers accordingly.  Cost: none; it
> rb> is trivial to implement in the lexer.  Benefit: it saves the
> rb> programmer from wasting time on spurious errors.  Finally,
> rb> any compiler that does allow nested comments will have a
> rb> competitive advantage over one that doesn't.
> rb>
> rb> Also, I feel strongly that the C defect of not specifying
> rb> argument evaluation order is not something that should be
> rb> carried over to this language.  I don't really care whether
> rb> the order is left-to-right or right-to-left, as long as we
> rb> have a standard.  I'm perfectly willing to take the extra
> rb> hour to rewrite my code so evaluation is left-to-right.  That
> rb> way we have a much greater chance that scripts will always
> rb> produce the intended results.
> 
> Wes Hardaker wrote:
> wh> I agree with you that this is a serious defect in C and I'd
> wh> like to not inherit C's bugs into this language just because
> wh> we're deriving from it.
> wh>
> wh> Actually, we're now deriving from C++.  I assume its not
> wh> specified there either?
> 
> If I could be convinced that either of these issues (1)
> are clearly standardized in ISO C or C++ or (2) are clearly
> set to be decided on in the Near Future for either C or C++,
> I'd be much more willing to back up these changes.
> 
> With kind regards,
> 
> --
> David Partain                  David.Partain@ericsson.com
> Ericsson Radio Systems AB      Tel:    +46 13 28 41 44
> Research and Innovation        Fax:    +46 13 28 75 67
> P.O. Box 1248
> SE-581 12  Linköping, Sweden