Re: Servlet based implementation of brl.

[Per Bothner <per at bothner dot com>]

On 08/17/2009 07:00 AM, Bruce Lewis wrote:
> I'm comfortable with moving gnu.brl into gnu.kawa.brl.

Or perhaps vice versa.  More BRL code probably references
gnu.brl than gnu.kawa.brl.

Also, gnu.brl is more similar to what I've some for other
languages (like gnu.xquery.*), and make less-verbose
sub-packages (like gnu.brl.util).

Give than you already control gnu.brl, I'm fine
with whatever you prefer.

> I think of BRL as an extension of Scheme, so it would make sense to
> share a namespace if that's practical.

The namespace of pre-defined names is clearly different.
Kawa does support a shared (multi-language) user environment.
The idea is that use user-context has its own user Environment.
By default this is per-thread, which means per-CallContext.
but it is possible to set the user Environment of a CallContext,
and thus multiple threads can share a user Environment, or you
can explicitly switch user Environments.  Each user Environment
is backed up by the "global Environment", which normally
is a shared instance of BuiltinEnvironment, which indirects
to the per-language builtin Environment.

See gnu.mapping.CallContext.getEnvironment and
gnu.expr.BuiltinEnvironment.  The Environment.global is
normally the BuiltinEnvironment.instance.

So a shared user Environment should be fine even if we mix
multiple Languages.

> In BRL 2 each page had its own scope.  You could define variables in
> individual pages without worrying about whether a global variable had
> the same name.  Does it work the same way in the Kawa servlet's
> module-based system?

The Kawa servlet interface could do with some TLC.  (One of my goals
is to add support for the com.sun.net.httpserver in Java 6.)
I had plans to dig into this - but after release Kawa 1.10,
and after spending some more time on performance issues.  But if
if you have some changes you'd like I'll certainly consider them.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/