Re: release methodology

[Roberto Ierusalimschy <roberto@...>]

> What about settling for something like the usual alfa/beta/rc/gold
> cycles (Tecgraf could use whatever nomenclature fancies them), where
> [...]

We try to use quite similar rules, but with different names. What you
call "alfa" we call "work", your beta is our alpha, your "rc" is our
beta, and "gold" is our "official" release. (Except that in no release
we have "bugs expected and accepted" ;-)

For sure "lua-4.0-update" should be called "lua-4.0.1-update" (the
"update" is not part of the name of the release, but the type of the
file. It is not a whole distribution). Our fault.

About the "gap" from 4.1w4 to 5.0w0: During the development of 4.1
several members of the list suggested (and we agreed) that the set of new
features "implies conceptual and significant changes in the language
specification", so it would be better to get a new major number. So,
what should we do? We were in 4.1w4, going toward 5.0. Somewhere there
would be a gap.


> Back to release methodology in general, another issue to consider
> is that some Lua developments (full lexical scoping, weak tables)
> are clearly useful and have been working well in the development
> releases for some time, while others (global keyword, generators
> library reorganization) are taking a while to get a handle on. It
> would be nice if things could occasionally propagate from development
> to production without waiting until all the stars are correctly
> aligned. That's more distraction for Lua's small team however, so I
> can understand why they may not want to do that.

We are also trying to make Lua more stable. Full lexical scoping,
although clearly useful and working well, is a source of incompatibility
with Lua 4.0. The same is true for boolean types. We didn't want to
release several new versions, each incompatible with the previous one.

-- Roberto