Re: C++ && squid 3.0

----- Original Message -----
From: "Alex Rousskov" <rousskov@measurement-factory.com>
To: "Jon Kay" <jkay@pushcache.com>
Cc: <squid-dev@squid-cache.org>
Sent: Thursday, March 22, 2001 11:39 AM
Subject: Re: C++ && squid 3.0

> On Wed, 21 Mar 2001, Jon Kay wrote:
>
> > Actually, I would suggest that C++ migration for the sake of C++
> > migration is not worthwhile.
>
> I agree.
>
> > 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.
>
> Hmm... To me, this approach is exactly what ``C++ migration for the
> sake of C++ migration'' is! :)
>
> To get the advantages of C++, one has to design for C++. Squid 2.x is
> not designed that way so just "moving things to C++" within each
> module is probably not worthwhile. Changing the design one module at a
> time does not sound like the right approach either because the
> "interesting" part is changing how different modules integrate and
> [re]use each other (i.e., "the big picture"). Just rewriting the
> interfaces in C++ is trivial (due to current Squid design) and, again,
> probably not worthwhile.
>
> These are just my very subjective opinions though. Since I doubt I
> will be able to contribute code directly in the foreseeable future, I
> will shut up.
>
> Alex.
>
>

Alex, I agree with you - to a certain extent. The authentication code
woulld have been a _lot_ easier if I could derive classes for
ntlm/basic/digest. My point is that a rewrite - doing it all at once -
while it allows maximum opportunity to work on the real benfits - the
big picture - restricts us now. And that the freedom to start doing the
bits of squid that make C++ sense in C++ would be useful. The tie into
squid 3.0 is that I really don't want to do (whatever bits I end up
doing) twice - once now in C++ because it made sense, and then rewriting
into C++ because we're starting over...

I guess I probably wasn't saying that clearly though.

Rob
Received on Wed Mar 21 2001 - 17:44:39 MST