[SNG-chat] Let's start something up (was: Re: [epona] Epona "is this the end")

[Daniele N. (J. <jo...@so...>]

(from the epona ml)

Gioved=EC, 23 gen 2003, alle 18:31 Europe/Rome, Lucas Nussbaum ha =
scritto:

> nobody with enough coding
> skills  willing to take the role of project leader. All the potential
> developpers are looking at each other waiting and thinking "I'll let
> them start, then I'll help". =3D> many super-contributors, but no =
coding
> team.

So, here we are again.
I think that the main reason for people not starting anything (apart=20
from a possible lack of time) is that no one knows exactly what=20
language this is going to use.

Lucas proposed Java, others didn't agree. His arguments about Java=20
being currently faster than the past, and machine-compilable, just=20
didn't convince others (I must admit I'm personally not sure).
So, I'm going to give my usual two cents about this thing, hoping that=20=

others will at least contribute in this discussion, which, we hope, is=20=

going to decide what language this stuff is going to be written in.=20
Otherwise I'll start doing something in Python myself. :)

---8<---
** Why Java would be good for services **
- Fast to develop
- Easy to mantain
- Compilable into machine code to run as fast as C stuff
- Runnable through a Java Virtual Machine

** Why Java would not be good for services **
- It could probably be slower than C
- Installing a recent JRE to make services live in a fast JVM would=20
probably be too tricky, if not impossible, in some configurations
- Installing gcc 3.x to compile Java stuff into native machine code=20
would probably be even worse

This is because many machines running irc services are often used for=20
other stuff, and they just can't afford the risk of upgrading the JRE=20
(they could be using java for some stuff that can't be interrupted), or=20=

even worse, upgrading to gcc 3.x to compile services. Upgrading gcc=20
could even mean, in the most Armageddon-like situation, that the whole=20=

system ends up being messed up by accidental overwriting of libraries.=20=

It's a possibility, although it's improbable. But it's still possible.

** Alternatives **
The most praised alternative to Java would be C++, even though C++=20
would be a PITA(TM). It inherits all the bad things of C, multiplied by=20=

the tricky use of classes. This means that coders would lose=20
significantly more time playing around with pointers and references=20
than writing real code. This isn't very productive, but could be eased=20=

by writing a good interface base to the internal services system. Maybe=20=

C would make servicesng a bit easier to develop than C++.

Other possible alternaives that I could think of would be interpreted=20
languages, such as Perl of Python, but I'm not taking them into the=20
discussion because most people would think that they wouldn't fit. I=20
mean, they'd probably work fine anyway, since we're not computing prime=20=

numbers or something like that, but the only way to see how good (or=20
bad) they scale is just a matter of checking AFTER the whole thing is=20
done.

Anyway, between Perl and Python I'd choose Python without a second=20
thought. It's easier to read, to develop and to mantain, and I could=20
help myself. And we could write the killer functions in C for the sake=20=

of speed.

Anyone interested, have a look at=20
<http://www.linuxjournal.com/article.php?sid=3D3882>.
--->8---

Regards
--=20
Daniele Nicolucci (Jollino) - jo...@so...
IRC Operator on Discussioni.Org (www.discussioni.org)