lua-users home
lua-l archive

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

On Fri, Apr 21, 2017 at 2:50 AM, steve donovan
<> wrote:
> On Thu, Apr 20, 2017 at 6:37 PM, Andrew Starks <> wrote:
>> layers: some kind of foundation uses whatever the authors of Lua
>> publish as the language that is the basis for an ecosystem, which
>> includes everything needed to build software systems and applications
>> with Lua and a host of curated libraries in the middle.
> It would be nice, but does it need organization? As Dirk says, any
> person with strong opinions and some reputation can post a set of
> recommendations. No blessing required! (It's not a bad idea actually,
> but who is going to bell the cat?)
> Part of the problem is the concept of 'nice'.  We have had some very
> heated discussions on lua-l, but *no-one* criticizes another's work.
> That's just not done, and that's cool. We value niceness.
> With a plethora of solutions, a kind of natural selection happens, but
> the painful process of discovery (as Dirk has gone through) is not
> sufficiently public. We need to learn from others' experience more.
> Look, *every* software ecosystem has this problem - there are nearly
> 9,000 'crates' for Rust, and finding your particular needle in that
> haystack isn't easy (even though actually using crates is a pleasure).
> So the 'discoverability' word comes up often. It works like this: you
> hang around with the people from a community, listen to their chatter,
> and find what's the canonical way of doing things. This is not
> particularly efficient, although very social.
> As for quality - oy, where to start? Documentation is an issue dear to
> me, but even if the standard toolkit provides easy doc solutions (as
> it is with Go or Rust) where you just have to throw some comments
> before your functions, then people will not take the hint. But as lhf
> says, documentation is hard, it isn't just about the tools.  Most
> module authors would far rather write code than documentation -
> open-source software is about maximizing fun, not work. (If something
> is fun, it is not work)
> We actually had a sweet spot with Lua for Windows, but it had a big
> problem: it was hand-curated, not build-from-scratch. Peter Drahos'
> LuaDist tackled that problem in an ambitious way, and provides
> 'Batteries'.  Which could become the base distribution, with LuaRocks
> for the rest, but (as Andrew and Ryan) know, it's a lot of work.
> So, the solution is not more structure, nor blessings from high - it
> comes from the community, which is to say, the efforts of those who
> think it is worthwhile.
> steve d.


I think that you're making a great case for the need for an
organizational structure and then concluding the opposite.

"Nice" or "polite" is the way things get done because it keeps our
emotions in check. I believe that nobody feels qualified to make
decisions because nobody wants to claim to have any authority. Peter
Drahos doesn't want to decide that luaPosix makes it in but cqueues
does not. Most people with the talent for software development don't
want to stand in judgement over other developer's decisions, even when
doing so would provide benefits.

A polite organization can do those things on the authority of the
people that vote the decision makers in.

Also, such an organizing body does not need to exclude any modules. It
could for example say, "These modules are tested and built by the
organization, conform to its highest standard, do not overlap each
other, work well together, reasonably follow a specific style and are
thus a part of a core set. Other modules are not built by the
organization and may choose to follow these standards or not and here
is a handy search mechanism for you to use to find them."

Also, I believe that LuaRocks and LuaDist are a part of the solution.

Andrew Starks