lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index] 

 > First of all, sorry about editing the thread subject yet again, but I
 > would never want to get into a "Rocks vs. anything" situation. :-)

Good change.  My bad.

 > The problem of course being that the Lua community (like any
 > programming language community) consists of people who create
 > software, unlike Linux distro communities, which have many
 > non-programmers.

I think you'll find the more Lua succeeds the fewer members of the
community create software that is reusable by others.  Years ago I
read an IBM document that says if a program works for you, making it
work for someone else will require triple your original effort, and
making it a product will require triple that effort.  I've never done
a product but have released plenty of stuff open source and that
factor of three is a pretty good estimate.

I'm wandering a bit (friends don't let friends send email after 1:00am)
but my point is we hope for a 'success disaster' in which a fraction
of Lua users write programs that are packaged and the rest want things
to 'just work'.


 > So, we're actually starting to see this packaging community
 > start to form.

Terrific!

 > Of course, the larger the overlap between the developer and the
 > packager community, the better.

I'm not sure I agree.  Developing good software is hard!  Example:
Fidelis Assis and I have spent over a year trying to turn his OSBF-Lua
spam filter into a reusable library that can be called by any
mail-handling application, and right now I wouldn't wish our code on a
dirty dog.  It will take months of effort to make it simple enough to
be good.

Packaging should be much easier!  If packaging is as hard as
developing we're all in big trouble.  So I hope there will be lots of
packagers who are not developers.  (By which I mean they do not
develop code for release to others.)

 > >   > > Furthermore, if one happens to be a debian user/developer, debs are
 > >   > > often more desirable for various reasons, so effort spent making them
 > >   > > is, at least, not wasted.
 > >   > and if one is not, these efforts are wasted. but everyone on every
 > >   > distro can install rocks and use it.
 > >
 > >  Turn this around: I'm a Debian user, so I can use a .deb for every
 > >  program written in every language.  Rocks are only good for Lua
 > >  programs.  Lua is a fraction of my output and a smaller fraction of my
 > >  research group's output.  So if I have time to learn only one, I will
 > >  learn Debian, because I can use it for all the software I write and
 > >  all the software my students write, not just software written in Lua.
 > 
 > Personally, I find both points of view valid. 

Oh yes.  We're in violent agreement!   The situations are entirely
comparable: you can get reuse across distros (rocks) or across
languages (debs) and either way you limit your target audience (which
is what I think someone meant by 'wasted effort'.

 > Returning to the developer/packager distinction, I think that from the
 > developer side, the most globally helpful course of action is to
 > provide good Makefiles, that will work nicely for all packagers. I
 > wrote about the main issues...

They are nice notes.  Make in general is an unbelievable tar pit
(catch Stu Feldman sometime and ask for stories).  GNU Make adds
staggering complexity.

I'm with the other poster who wished it be possible to create at least
a simple rock without any makefile at all.


Norman