Re: Minutes of the GNOME Board meeting 10 July 2001

[Havoc Pennington <hp redhat com>]

linas linas org (Linas Vepstas) writes: 
> Not to waste your time, but ... what are the various lines of thinking?
> (I'm guessing, but 
> -- one must be to build MS-compatible tools; that seems to be de Icaza's 
>    tack; 
> -- the other must be along the lines of 'let the smartest volunteer
>    coders build what they feel like, for verily, they will make it
>    technically correct and therefore superior.'

Well, there are several levels of problem to address. One level is
simply to be able to deploy chunks of code in a way that's dynamically
introspectable and implementation-language-neutral. This can be done
via a native code component system, or via a rich language runtime
such as Java or the .NET CLR. Next you need to export your interfaces
via remote protocols; this comes nearly free from the dynamic
introspection. Another issue is how to ship code around between
machines in a secure fashion; this requires the concept of "safe
bytecode" as with Java or .NET. Also there's the matter of
packaging/deployment, which .NET defines much more thoroughly than
Java. Then you have to actually define the interfaces, protocols, and
XML formats that will be used to link various chunks of things
together. And we have the whole peer-to-peer area where things are
still very much in an experimental state, no one has gotten it to
really work well. These are all big hard problems.

I believe the main thing we should do in GNOME is work with broader
open source efforts to address these things. I would expect many such
efforts to exist, and I would work with as many of them as we can.

In my opinion native code introspectable components are a useful,
easy-to-implement thing to go ahead and have. But you need some more
elaborate language runtime eventually. If Java were open source it
would be far more interesting in this respect and I think the .NET CLR
would have no chance of usurping it. Also Java needs to work on
packaging/deployment and support for multiple language syntaxes, but
that could be done. Unfortunately none of the companies with decent
Java implementations have seen the light on either the licensing or
the technology, at least not that I know of. I remain hopeful
nonetheless.
 
That all just tackles the first few levels of the issue though; for
peer-to-peer and so on, things are too experimental to even speculate.

GNOME should really remain agnostic, and see how things play out. We
have no reason to commit to one technology or another in the short
term. As technologies become useful, particular pieces of GNOME are
welcome to use them, and we'll have a de facto decision eventually.

GNOME is a deployed end-user application, and I think it is nearly
always right for such applications to be conservative about adopting
new technologies.

> I presume the important discussion happens on the gnome-hackers
> list?

Most of the GNOME lists are focused on making a desktop. ;-)

These big "future of computing" meta-questions tend to come up mostly
in casual conversation or whatever. There isn't an official GNOME view
or decision, and IMO there shouldn't be, certainly not at this stage.

Our main challenge right now is to execute well - to deliver good
desktop technology on a regular schedule. For these bigger issues, I
think our responsibility is to think about it and be an interested
participant as the whole open source community tries to address them.

Havoc