Re: Emacs rewrite in a maintainable language

[Marcus Harnisch]

Daniel Colascione <address@hidden> writes:

> We can
> always do better with application-specific data structures than we can
> with one-size-fits-all C++ data structures designed as reasonable
> defaults, not optimal implementations.

With all due respect, that is a weak argument[1]. Having things under
your own control does not guarantee better (see below) performance. At
best it may carry the potential of delivering better performance.

> I'd hate for someone to think
> that std::unordered_map, say, is somehow better than the data structures
> we've used for a long time merely because the standard stipulates that
> the runtime must provide it.

I guess that depends on the definition of “better”. Regarding run-time
and/or memory usage, it is much more likely that, say, a processor
vendor will contribute a clever, architecture specific implementation
to a compiler's standard library than to Emacsen.

But then I was just raising this point as another potential in the
unlikely event of someone converting the code base to C++, and not in
order to convince anyone.

Regards
Marcus

Footnotes: 
[1] Just as weak as my own attempt to paraphrase a comment whose
source I have been unable to locate all morning :(