Re: comment about gnome architecture

[Danilo Segan <danilo gnome org>]

Sorry to jump into the discussion, but I cannot resist temptation :)

Greg Breland <gbreland mozillanews org> writes:
>
> Its much easier to transition 12 applications that implemented a bad
> spec to a new better version of said spec than to move 12 applications
> that did it their own way to a good spec.  
>

What are you basing your conclusion on?

I can imagine a situation where that would be the case, but I can also
imagine a situation where that would *not* be the case (eg. change
from bad spec to good one involves completely different mechanisms,
and thereby requires rearchitecting entire code-base -- this means
having to change *all* 12 apps that conformed to bad specs, instead
of changing 5 or 7 which used completely different abstractions).

I would never be as brave as to claim that any one of the ways is
guaranteed to give better results. There are many more factors, and
it's not as simple as that. Sometimes are more people attracted to a
specification if it is presented in a form of "prototype that works"
rather than by "specification nobody knows if it will work".

> I would also like to assert that the earlier a spec is written the
> better.  Writing a spec after someone notices that 4 programs are doing
> xyz in their own way is too late.  Of course, this almost guarantees
> that the spec won't be very good in its first draft form.  The
> alternative is that there will always be some percentage of apps that
> can't be bothered to change something that already works.

As far as I see, good code is specification no other can match. Yeah,
sometime hard to follow, but many specs are ;)  What's else, I think
nicely written and understandable code is to be preffered over bad
specs.

Now, to the _original_ topic with my overly long jibberish ;)

My feeling is that Gnome is all about infrastructure, because if it
weren't, nothing would work. All the stuff that may seem 'obvious' to
outsiders is actually making use of lots of infrastructure, and
that's what makes it tick -- infrastructure.

A pure doubt about Gnome infrastructure original poster expressed,
also points out that whatever infrastructure is there is well
abstracted, and doesn't interfere with common usage. This is where
gnome-vfs, mime databases, .desktop files, a11y, intltool, XML, gconf,
glib, bonobo and CORBA all fit in; if that's not infrastructure, I
don't know what is.

Simply a lack of one accepted and shared specification (which, btw, is
lacking on any other system I know of) -- that of bookmarks -- is not
something to base final conclusions on.

Like always, there's room for improvement, but lets not diminish what
is already present. And whoever has had to create any app on par with
Gnome apps without using any of the mentioned tools knows how much
does Gnome infrastructure really mean.

Also, improvement is happening.  In just the past month or so, we
were greeted with several infrastructural proposals for Gnome.  I'll 
pop two off the top of my mind-stack: gnome-keyring and
evolution-data-server. 

The real problem I see with all this is it not being advertised well
enough, and not all potential developers are tracking
desktop-devel-list.  With Gnome Summaries reinstated, I hope this will
get better too. 

Cheers,
Danilo