World: r3wp

No GPL derivitive except Classpath or (L)GPL3 can be encapped, for 
instance.

Legally, I mean. You can't encap (L)GPL 2 code.

Extensions are compatible with LGPL, but not statically linked or 
bundled code.

Too bad - Boron sounded promising.

their argument will be that its rebol that is evil, cause its partially 
closed.   but right now I don't really care.. there are a lot of 
nice things comming for R3.

If Boron changed to Classpath or BSD/MIT then there would be no division 
of labor between the Boron and REBOL communities.

The R3 host isn' t dynamically linked? In the original plan, we were 
promised both static and dynamic libraries

There' s no problem with encapping LGPL code. You just have to deliver 
the object files that allow the receiver to reproduce the encapping

The host is currently statically linked to the kernel (afaik). Most 
host builds will be statically linked in any case.

That would be very bad, not just license wise, but also in terms 
of system performance

Why? Dynamic-linked function calls are slower.

A little, but you get to load the entire environment over and over 
again for each, possibly short-lived, REBOL process you start

This eats memory and startup and teardown time

If the shared library is withdrawn, that' s very bad news

Do you realize that most of R3 will be open source? The static linking 
is just for performance.

I was just hoping that Boron would choose a permissive license instead 
of a divisive one.

I am well aware of the situation, except for the parts that we have 
been unable to know in all those years, such as the eventual license 
and software configuration

I will not get into the anti-divisive properties of the LGPL and 
GPL here

I was just trying to figure out a way to endorse Boron and say that 
it is good for the REBOL community. Sorry.

It was originally thought that the kernel would be dynamically linked, 
but the performance of that was so bad that static linking was the 
way to go. It will still be refined in any case.

A large portion of the code in R3 will be in the host code, so having 
a dynamic linking break there won't give you as much benefit as simply 
marking pages as sharable or something.

Is that official?

When you consider that all of the platform-specific code is in the 
host, it's obvious.

No, I have always planned on the basis of a shared library, which 
is standard practice and was promised

I' ve been trying to do the very same thing defending REBOL in all 
those years, for example in the Syllable project. It's very hard

If it turns out that more broken promises make R3 unusable for me, 
my only saviour will be Boron

I will have been made to wait for half a decade for nothing

We'll see. I didn't know that Syllable was open-source only - I was 
keeping it in mind as a platform to be supported. By hybrid-source 
builds, but still a planned target platform. I'm sure having Boron 
as a R3 kernel replacement would be possible, as long as it is license 
compatible.

Syllable is an open source project and was always clearly presented 
as such. We do that for one overriding reason only: to never get 
in the Atari/Amiga/RiscOS/BeOS situation again, where commercial 
entities destroy your platform

Kaj, do you realize that the entire host and kernel combination could 
be a shared library? That would solve your startup problems without 
the performance hit. Or you could split your host into platform-abstraction 
and platform-integration portions and then dynamically link between 
those parts. It's just putting the split between the host code and 
the kernel that doesn't make sense.

I know that Syllable is itself an open source project, but I thought 
that you would allow close-source applications to run on it. Especially 
freely distributable ones.

Yes, applications. By considering closed system components I am treading 
a very fine line. We can never make the base system dependent on 
closed components, for the very reasons we are discussing now

I always presented REBOL as our cross-platform strategy. As such, 
I have defended part of it being closed

The host-as-plugin-for-other-app is one of the models that R3 was 
designed for. That means dynamic linking.

Then why do you say there will be no dynamic library?

I say that there will be no dynamic split between the host and kernel, 
for practical reasons (performance drops like a stone). I don't say 
that you can't make a statically linked host and kernel into a dynamic 
library that other things can link to.

There is no problem with replacing the closed portions of R3 with 
open source clones, afaik. Except license incompatibility of course.

kaj, the main reason for the close source is to prevent as much of 
linux-like split as possible to the *Language*.

the platform is a totally independent aspect of REBOL.

It' s up to Carl to choose a license that will either unify or divide 
efforts

But we'll find out for sure soon, don't worry.

I know the philosophical underpinnings of REBOL, but we also know 
where they have gotten us so far

In Syllable, we are building a unified system out of open source 
components

afaik, Carl wishes the core to be linked in any situation.  There 
are still a few things I'd like in the core to migrate to host, but 
as I know Carl, factual experimentation has more weight than theoretical 
ideas.

Kaj, there are certain licenses that can't be chosen because of things 
like this. *GPL is one of these. Parts of R3 will be closed source, 
so licenses that reject that won't be compatible. If you have a problem 
with that, ask Karl to relicense Boron as Classpath.

so once we have the host and Carl realized that he'd waste less time 
giving us a bit more control, there is a chance for a bit more core->host 
migration still.

Maxim, yes, it would be a great surprise to me if Carl would put 
a bomb under the linking abilities

he just want to protect the language so that we don't end up with 
things like a version of REBOL with commas everywhere... ;-)

... the horror :( ...

I thought that' s called a dialect? :-)

hehe... a dialect still has to obey the syntax rules... commas are 
specifically designed *out* of the language.