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

Re: current state of green on windows

Jeff Rush wrote:

> roger gregory wrote:
> >
> > I've got Green compiling on windows under the current cygwin compiler.
> FYI, I'm deep into the process of cleaning up the Green sources, adding
> ANSI C constructs, cranking up warning levels and fixing compiler
> warnings
> and such.  I am then going to restructure the source, regrouping
> functions
> from the functional arrangement they are now, to a more object-oriented
> layout.  This way people can extract just the parts they want -- such as
> the crum subsystem, or an enfilade class object.  I'm also adding
> distinct
> interfaces btw the FEBE protocol, the indexing/linking engine, and the
> backend storage management, so that any can be swapped out with new
> designs (e.g. integrating the Ent from Gold or using Freenet for the
> backend storage) or recycled into new projects.

Well theories a lot there to reply to.  The ent isn't (may not be ) generally superior,
the indexing stuff needs to be fixed, especially to allow protections, and editing
of links (this requires changing the spanf in the current implementation), and a lot
of other things.

> I will be packaging the results as Red Hat Linux RPMs as well as your
> basic tarball, hopefully within next week for the first pass.  There
> will be a 'backend' RPM and several 'frontend' RPMs.
> > However I'm having problems with pyxi, the line x88.py:899 ( in my version ) is
> > os.mkfifo(self.fifo)
> > which is a unix only thing.  I could dig up the old way to do it, like we did in the
> > released c frontend, but I'm more profitably  engaged in digging up documentation.
> Just wait a bit -- I'm fixing/packaging pyxi to run via the more
> commonly
> available TCP stream to a remote server, instead of spawning the backend
> and
> communicating via a pipe.

If you look, Pyxi will currently do this as well.

> > I've got Gold documentation of about 5 Mb, I'll be putting on the web as  soon as
> > I clean it up (from Frame ca 1992 Mac via Tops mount, if you know how to convert
> > it contact me).  I can send the tarball if you want it before then.
> Send me the tarball -- 5 MB is no problem for my inbox and I will make
> use of it.
> > I've also got about 5 Mb of mail tarball archive, that I need to scan for actionable
> > stuff before I can post it.
> >
> > I'm pushing on Dean Tribble to check on Parc Place smalltalk status, so I can publish all
> > the Gold stuff.  I have a source of Sun IPX machines at $39.95 +tax with 64M ram and
> > ~300M disk no keyboard, mouse or screen, that will run Sunos4.1.4 and the ParcPlace
> > Smalltalk.  Contact me, if you want to get a copy of the entire thing, we have a limited
> > number of licenses and can justify a few copies till we get clearance to give
> > ParcPlace Smalltalk 2.4    away (yes I know ParcPlace got sold but, till I know that
> > the rights are pd, I can't publish it).
> Sounds good, but for my needs I'm not looking to run SmallTalk but to
> extract the
> interesting portions of Gold such as the Ent and back-fit them into
> Green.  I'm
> learning SmallTalk syntax right now, and dissecting Gold.

Threre's a lot more to gold and than what's currently distributed, the C++ source code exists
but is maintainable only via our LARGE set of tools and the Smalltalk/C++ translator.  One of the
things that is interesting is the idea of taking the Smalltalk environment and having it generate
Java as well as/instead of C++, or porting the environment to Java, or some other language.

One problem with Java is that since the byte code interpreter doesn't have the type info
they are limited to crude garbage collectors, like the Train Algorithm.  More on that later...

> -Jeff Rush

 Roger Gregory
 541a Presidio Blvd
 The Presidio
 San Francisco ca 94129
 415 292-6071

the software but not the name!