Re: C++ && squid 3.0

Alex Rousskov wrote:
>
> On Wed, 21 Mar 2001, Robert Collins wrote:
> > On the squid-3 page it suggests using C++. I've got no issue with that . . .
>
> IMHO, C++ code is not just a better C code. Simply converting existing
> C code so that it compiles with g++ is possible, may be even
> beneficial, but is probably not worth the effort. To gain the benefits
> of C++, a significant change of Squid core interfaces is unavoidable.
>
> Yes, there still will be modules and such, but the relationship
> between related modules and code sharing principles are likely to
> change a lot.
>
> If I have to decide now, I would propose two semi-independent C++
> migration activities:
> - conversion of basic data types to full-featured classes
> - overall design changes and converting modules to classes
> and algorithms
>
> There is no doubt, however, that most of the functional code within
> the modules will be preserved almost as is (syntax-like changes are
> trivial and not important).

Actually, I would suggest that C++ migration for the sake of C++ migration
is not worthwhile. The advantages are offset by the temporary surge in bugs
and the fact that the rewrite is probably done by somebody with imperfect
understanding of the module rewritten.

The best approach, IMHO, is to change each module to C++ as each module sees serious
change -e.g., the person making significant algorithmic changes gets to move things
to C++. That minimizes the negative impact, since it means that people who understand
each module will do the job on that module.

                                                        Jon

The advantages will be over
>
> Alex.
Received on Wed Mar 21 2001 - 16:47:25 MST