Re: Heated agreement? (was) Re: Canvas shortcomings

[Mark <jamess1 wwnet net>]

On 30 Jun 2001, Lauris Kaplinski wrote:

> Hello!
> 
> I see the temperature raising...
> 
> There is problem in designing generic interfaces for gnome, yes. But
> I see no easy solution in current development model. And I am not
> very sure, whether changing development model makes sense.
> 
> Each maintainer is genuinely interested in his/hers project.

I respect their decision, as I had no part in the development.

> I am quite
> sure many people are capable of writing decent API abstractions, but
> alone these do not make sense. And each project implementing its own
> slightly incompatible abstraction layer does not make the system
> better - even worse - instead of good specific API-s, we will have
> something like:
> gnome-print API -> canvas API (print preview)
> canvas API -> gnome-print (adding ::print to canvas model)
> canvas API -> display abstraction API (proposed Gdk/libart/gp commonAPI)
> display abstraction API -> gnome-print API
> and so on...
> 
> So IMHO generalizations have to be addressed in separate layer,
> developed mostly independently of underlying projects. Anybody can
> start such project - and if it turns out to be good, I am quite sure
> it will find use(r)s.
> 
> IMHO gnome has deliberately chosen an approach to build loosely tied
> framework of independent technologies. Whether it is the right decision
> in long run, or whether it will ultimately be superceded by something
> gnome-VM like, we will see. AFAIK KDE (Qt) has chosen more VM-like
> approach - which is logical, as its core is basically evolved from
> single project of single company.

VM = Virtual Machine?

It seems any project sort of targeted to a VM, since PC's and workstations
in general are very heterogenous.

I agree with the aproach of developing libraries with very specific
functionality. In general I think this good for free software, in
ecourages a clean design and reuse in other projects. Though, from my
experience, it seems Gnome has many mixtures between techonology, guis,
abstract interfaces. I don't think this is a good design stratagey, these
three should seperate and distinct.

Design-wise, I think Gnome should still specify an abstract interface for
the target development platform. The Java API in particular is an example
of a very coherent and consistent development platform, although 
techincally I think a better platform can be designed. Implementation wise
Java (and maybe Qt/KDE?) is very monolithic. I agree free software doesn't
tend to be developed in such a way. However, what Gnome can do is specify
a target design, and use what ever free software is available to implement
this design. Sometimes this might spawn new projects, and I would
encourage such projects to be very specific in their functionality, but I
realise I have no such control over what people would like to develop.

To sum up, I think Gnome would benifit from somewhat of a design oriented
aproach. It seems at some level development ideology (Bonobo) takes
precedence over design and technological considerations.

Mark