2.1 vs. 2.2

[Martin v. Loewis]

Alex Martelli <aleax at aleax.it> writes:

> Not what I saw happen in the small set of languages whose standardization
> I had to follow.  With C++, in particular, the standard was so far from
> "allowing all variations that already existed" that, AFAIK, still today, 
> years later, no commercial implementation claims 100% standard compliance
> (something that I still find rather amazing).  Of course, the usefulness
> of that is as dubious as that of a standard which did allow everything.

As an co-author of a C++ compiler, I can assert that the C++ standard
very useful both to C++ users and C++ vendors. It is easy to analyse
g++ bug reports against the standard; it turns out that you can close
many reports as "not a bug" if you have the backup of the
standard. Users agree that they should not dispute the standard text
with their vendor (in the rare cases where they dispute the conclusion
that their code is wrong, they usually agree that they need to talk to
the committee, reporting defects etc).

For this kind of application of a standard, it is immaterial whether
g++ claims to implement the standard "in full". There are obvious
areas where g++ intends to comply, and deviations in that area are
considered as bugs, and there are areas where g++ admittedly (and in a
documented manner) omits features from the standard. Users generally
appreciate that distinction, and try to avoid unimplemented features,
or use official work-arounds.

Regards,
Martin