What madness! Everytime I think of a reason to do something in C/C++, I find a simpler (and quicker) way to do it in Tcl.
This is hurting my (future) employability. In the past, I just watched the C++ coders slowly grind through getting an app up and running, while I secretly implemented an alternative solution quickly in Tcl... I can't bring myself to working (exclusively) in C/C++ again.
Standard C++ has a lot of good things going for it (its not your father's C++), but I find myself constantly re-implementing nice to haves (like fileevent, glob, file, etc) and quickly come to the conclusion that it won't be anywhere near as stable as the Tcl Core implementations --- so I go back to Tcl.
In any I/O bound application (i.e. bound by human interaction, internet connections, etc), I am hardpressed to find C or C++ outperforming Tcl. Or, to put another way, I can send 3-5 emails per second using a pure Tcl SMTP client (homegrown) in one thread of execution. The internet and SMTP is slowing me down. A similiar client in C/C++ is no quicker (and using Java + JavaMail is actually slower).
Tcl + Unix + (insert Tcl bindings to C packages here) = a complete application framework.
In particular, I just implemented a Tuplespace in Itcl (+ Metakit for persistence) in just over 350 lines of code. It performs beautifully (you can even use telnet to attach to it and put/read/take items). I dropped plans to rewrite it in C++.
-- Todd Coram ranting and raving about not much of nuthin
Bravo, I can only nod in total agreement - and will add deployment hassles as a weak spot of C++ code. I can't wait to see the Critcl system mature just a tad further, and really kill the (large-scale) use of C++ and even C. -jcw
Another idea would be to generate C++ source code ... -jcw