Lua for MUD development was Re: [vworld-tech] Modern MUD server design

ceo ceo at
Wed Apr 27 05:44:24 PDT 2005

Brian Hook wrote:
 > A little followup now from my Lua notes.
 > ...and, this sucks. =)
 > ...and, this sucks too.
 >>There are many support libraries, however not nearly as much as
 >>Pike or Python or Ruby.  But if you just need SQL and sockets and
 >>basic things like that, those are all available.
 > The lack of up to date and somewhat sanctioned support libraries is a
 > serious issue for me.
 > Lua folk themselves, and this is a liability.
 >>Minor typos can mask bugs for a long, long time unless you have
 >>really good code coverage analysis happening.
 > All of the above really made Lua unfeasible when the code base got
 > beyond a few thousand lines.  Things got REAL rickety, and without a
 > proper debugger/editor (I use Eclipse + PyDev for Python) it just
 > wasn't tenable.

All of which I could have (possibly did) flagged before you started, 
having experienced similar several times with different languages, and 
still being bitter about it now (the disappointment after getting hopes 
high). I don't contest that there are langauges that are theoretically 
better suited to our domains than C++, java, php etc (the core of 
extremely heavily and widely used langs) - but the practicalities of 
these exotic languages tend to make them unusuable unless you enjoy 
fixing the language as you go along (tongue in cheek...).

 > So now I'm a Python whore, but I'm still just learning it.

So now I'm a Java whore, and spend much of my language-debate time 
explaining to people how well java can solve their problem NOT because 
it's "the best language for that problem" but because it's often 
"practically speaking the best fit, whilst being *good* for that problem 
(though others are clearly better)" (where C++ is often "practically 
speaking bearable but not great, although maybe better than java for the 
problem"). The whole "You can do anything in C++ ... but only if given 
infinite time and manpower" issue :(.

Interestingly, what I've seen *in practice* of Python strongly suggests 
it shares the problems you cited of Lua (to a lesser degree) and is also 
mediocre performance-wise. Largely from personal experience (with large 
platforms like Zope which require lots of python work to maintain - this 
gives you a good idea of how nasty a large python project can be without 
having to spend 5 years building one to that size, although of course I 
wouldn't base such conclusions on just one app :)).

Shrug. Most devs I've spoken to at the GDC etc who are getting hooked on 
Python (leaving aside those who've been using it for a long time 
already) seem to be doing so for specious reasons - like the fact that 
it has easier mem management than C++ (!), or that it's "just as fast as 
C++ and trivial to port to C++ if it's ever too slow". I'm still pushing 
ex colleagues and friends who use it in production to give me some solid 
reasons for using it, and still waiting for anything detailed and robust :(.

Likewise, incidentally, trying to get conclusive reasons for moving to 
C# at the moment, so I can evaluate it seriously. Like Python, it looks 
good from the outside, but I've seen just enough nastiness to be wary yet...

More information about the vworld-tech mailing list