Re: Lua in NetBSD

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Sat, Oct 24, 2009 at 05:28:58PM +0300, Antti Kantee wrote:
> 
> My objection was that the only thing written with Lua in NetBSD was
> emails.  Writing longer emails will not alter my stance.
> 
> I don't see what's so wrong with writing code for your dozens of use
> cases Lua is perfectly suited for and giving people the chance to
> actually evaluate your proposal.

I don't see why you propose this as the only option.  Let me be 100% clear
about why we (Coyote Point) will certainly not do this: we have limited
development resources which are already more than fully allocated.  We are
able to shift our development priorities only when we can make a compelling
case that doing so will somehow reduce our costs or increase our revenues.

Doing large projects in NetBSD under conditions such that they might then
be rejected by core, and if we'd relied on them in our own tree in the
meantime we'd be left alone to maintain them over time does not meet
that criterion!

We've done a lot of work on NetBSD in the past few years.  Some of it was
stuff others use (e.g. accept filters) and which was pretty easy to do.
Some of it was stuff only we seem to use (e.g. our extensive enhancements
to /dev/crypto) and which was also quite hard to do.  Unfortunately, in
that case in particular, others basically told us "make the code available
and we'll help with it".  That didn't happen.  So the supposed advantage
of open-sourcing that code in particular has not materialized for us.

If I'm going to devote very expensive engineering resources to a project
in NetBSD, it has to fall into one of the two intersections {easy, may
be adopted, used, and supported by others} or {hard, will be adopted, used,
and supported by others}.  Because we're not heading for {hard, we're on
our own} territory again.  Period.

And that means no integrating large things into NetBSD in the hopes that
after we've done so, maybe someone else will nod his head and say "great,
now that you did all the work just in case, now this is OK".  The risk
is too high and the other work we could be doing in our private tree is
too valuable.

I think we (and Marc, and Jason, and Lua's maintainers, and
many of the others who have come out in favor of Lua integration in this
discussion) have a pretty good record of getting most of the major things
done that we say we'll do.  I think we are all saying pretty much the
same thing: "Tell us the work won't be thrown away, and we'll get it done,
because we all have valuable uses for it".  I doubt any of us are willing
to trade that point of view for your "do the work before getting any kind
of approval of what you intend to do": again, the risk of wasting our time
is too high.

However, I have to say that my impression is that the majority of the
developers who've spoken up in this discussion see it more like our
way (which is to say, not "agree completely" but "are tending in that
direction") than your way.  Do you disagree?

Thor