Re: comment about gnome architecture

[Greg Breland <gbreland mozillanews org>]

On Sun, 2003-12-14 at 19:13, Maciej Stachowiak wrote:

> Focusing on "pure infrastructure" seems like a recipe for delivering a
> lot of shining pearls of theoretical perfection that no one wants or
> uses.
> 
> It's better to focus on concrete goals that directly impact the user,
> and let that drive the infrastructure. It's true that some kinds of
> problems can't be solved by working incrementally, but when making a
> major infrastructural change it's even more important to keep the end
> goal in mind.
> 

I agree with you, I guess in trying to be concise I didn't fully explain
my point.

Major infrastructure changes should be done with a firm understanding of
the end goal.  However, there are lots of small infrastructure pieces
that can be done without a clear need by anyone.

Maybe my definition of infrastructure is different from yours? 
Infrastructure doesn't have to involve one line of code, it can be a
pure idea.  For example, I consider the thumbnail spec on
freedesktop.org as the perfect example of this case.  It is just an 
idea of how and where to store image thumbnails.  If it is a "unused
pearl" no one is any worse than the previous placing images all over the
place. 

So say I create a really lame bookmark spec(which is likely since I
don't use bookmarks) that is really not well done, or not needed or
overly complex.  Maybe a couple of developers use it but the vast
majority don't.  It dies and we move on no worse off, except for my time
creating the spec.

I agree that we might generate lots of shinning pearls that no one
wants.  However we are always taking that risk since even if you create
something, the developers don't have to follow.  The trick is to create
it in such a way that even if it becomes deadwood, it has zero impact on
the overall system.

For instance, if the thumbnail spec doesn't "win", it will eventually be
removed from freedesktop.org or something will take its place.  If gconf
doesn't "win" then it will linger forever.  One is major infrastructure
and needs to be driven by the applications, the other is small
infrastructure, but both are very very important.  I don't think we have
to take the same development approach to both types.

I guess my view boils down to this.  A bad spec is better than no spec
at all.

Greg Breland


> People don't use Windows because it has "good infrastructure". (In
> fact, few people would even make the claim that it has good
> infrastructure.) They use it because it has the apps and features they
> want.

Again, I think we have a different definition of infrastructure.  I
consider specifications for how and where to store settings and data
infrastructure, which is what Gnome is missing.  The true big
infrastructure of Window is awful, but the smaller stuff is deep.