Re: GNUstep theming (was Re: Objective-C 2.0 and other new features in Leopard)

[Dr Tomaž Slivnik]

> But, as project maintainer, I'm sure you can appreciate my  
> position. I can't say unilaterally that I want to appeal to one  
> group over the other.
>
> GNUstep currently most appeals to former NeXT people who are into  
> Mac OS X. However, a lot of these people also say that it's time  
> for GNUstep to move forward with it's GUI look. A friend of mine  
> owns a software company that was once fairly well known in the NeXT  
> world and he's said the same thing.

I do appreciate your position. I also appreciate that it is your call  
to make!

However, even if some (or a lot of) people are saying that GnuStep  
must change its look and "move forward", that does not make it either  
the right technical decision or the right business decision. You put  
it well in another thread yourself:

> Because everyone else is doing it certainly is NOT a compelling  
> reason. I prefer to think for myself.


Furthermore it does not mean that changing the appearance of GNUstep  
will result in more people using GNUstep. Where is your business case  
for that - what evidence do you have that this will happen?

Changing the GUI will not come for free - there is the cost of work  
involved in implementing it and the much greater still cost of  
increased future work and complexity involved in maintaining several  
different looks and feels. Indeed, I'm willing to bet that not all  
themes will be properly maintained, that possibly none of them will  
be properly maintained, and that this will lead to many tedious my- 
app-doesn't-look-right-in-xyzzy-theme bugs that will take forever to  
fix.

There are also risks involved - you're more likely than not to make a  
dog's breakfast of it. The people who designed the NeXT look and feel  
were very good, and nobody has yet succeeded in bettering them. Apple  
tried, and failed spectacularly. "Modern" is not always better, and  
"moving forward" does not necessarily mean "improving". If you get it  
wrong, you'll end up not only not attracting new developers, you're  
also likely to drive away your core NeXT afficionados.

Worse still, you might get it "right" and attract lots of low quality  
developers of the sort who cannot see beyond screenshots and skins,  
who will fail to understand GNUstep in the same way that Apple  
engineers have failed to understand OpenStep. More developers is not  
always better for the platform - there is no platform in the world  
with more developers developing for it than Windows; NeXT had 500  
people working for them at their peak - who had the better software?  
Apple has 10,000 people - did they improve OpenStep, or make a dog's  
breakfast of it? What you want to attract is more *quality*  
developers. Trying to change the look and feel to attract the sort of  
developers who make their decisions as to whether to adopt a platform  
or not primarily on the basis of the look of its widgets and on  
screen shots is taking you into the domain of Apple-land, in my opinion.

These costs and risks do not mean you shouldn't go ahead with changes  
- but they do mean you need a pretty compelling business case that  
there are good reasons to believe (not just "some people think/say")  
that there is real potential substantial return to be had by pursuing  
this course. Who will switch to using GnuStep/developing for GnuStep  
because you've "modernized" its theme? What evidence do you (and  
others who claim that) have they will?

Mark Grice wrote:

> And, I still read that some people can't manage to get it  
> installed. This is a real problem.


Renauld Molla wrote:

> When you say that people interested about GNUstep won't care about  
> a new UI, this is definitely wrong. ... a friend of mine ... needed  
> an interface ... I advocated GNUstep ... why wasn't it retained?  
> Because for him, GNUstep is too hard to install ... and it looks  
> terrible. And this guy is no newbie. And there are loads of them.

Both people make precisely the right point although Renauld, in my  
view, mentions two problems and focuses on the wrong one. The reason  
GNUstep is not taking off, has nothing to do with its look and feel.  
Its look and feel is spectacularly good - why try to look more like  
everyone else when you already look better than them? You'll end up  
like Apple who has been trying to look more like Windows (very  
successfully) and has been getting worse and worse for it. There is  
no room for another Wintel/Apple clone in the market - look the same  
and it means certain death, look unique and different, and you might  
just find your niche.

The reason, in my opinion, that GNUstep is not taking off is  
blindingly obvious and much more mundane: the damn thing is  
IMPOSSIBLE to install. Who's going to use it if they can't install  
it? If you want to win the lottery, rather than worry what's wrong  
with you and "fix" what is already perfect, why don't you try buying  
a ticket?

Perhaps not on Linux, but on an Apple installing GnuStep is  
impossible. Why isn't there a .dmg installer somewhere with a .mpkg  
in it which just installs the thing on a Mac? Such a thing should  
contain all the development tools (gnu c with Gnu runtime, and all  
GnuStep software etc.), should run on any Mac OS X (10.3, 10.4, 10.5)  
and be fully multi-architecture (ppc,i386,x86_64 at a minimum). I  
just spent nearly 2 weeks trying to build GnuStep from scratch,  
hoping to produce such a .dmg, put it on my web site and offer it to  
you. I had to give up! I did not want to use DarwinPorts initially  
(for various reasons) but my only conclusion from that experience is  
that building the whole thing from scratch is just not a viable  
proposition. I'd built up all the prerequisites 5-way fat (ppc/ 
ppc7450/ppc64/i386/x86_64) only to realize that if I want to build  
GnuStep itself 5-way fat, I have to build 5 different versions of GCC  
and build everything else again with Gnu runtime ... and then rebuild  
GnuStep again. It's several more weeks' worth of work!

Eventually I got so exasperated trying to do it myself that I  
conceded defeat and tried the DarwinPorts route - on a brand new,  
cleanly installed machine, with just a /usr/local and an / 
Applications/Local added. First thing, DarwinPorts failed to install  
at all because it didn't like something in my /usr/local. Sigh, OK,  
so I archived /usr/local into a .dmg and removed it, just so I could  
try GNUstep. Try again! But even with a completely clean machine,  
DarwinPorts failed, like this:

> --->  Extracting gnustep-base
> --->  Configuring gnustep-base
> Error: Target org.macports.configure returned: configure failure:  
> shell command " cd "/opt/local/var/macports/build/ 
> _opt_local_var_macports_sources_rsync.macports.org_release_ports_gnust 
> ep_gnustep-base/work/gnustep-base-1.14.0" && ./configure --prefix=/ 
> opt/local CC=gcc-mp-4.2 GNUSTEP_MAKEFILES=/opt/local/share/GNUstep/ 
> Makefiles " returned error 1
> Command output: configure: error: cannot find install-sh or  
> install.sh in /opt/local/share/GNUstep/Makefiles "."//opt/local/ 
> share/GNUstep/Makefiles
>
> Error: The following dependencies failed to build: ArtResources  
> gnustep-core gnustep-back gnustep-gui gnustep-base gnutls libgcrypt  
> libgpg-error gettext libtasn1 lzo opencdk readline ncurses ncursesw  
> libpng libungif tiff jpeg libart_lgpl GMastermind GMines GNUMail  
> Etoile SQLClient Performance sqlite3 gawk dbus docbook-xml-4.1.2  
> xmlcatmgr xmlto docbook-xml-4.2 docbook-xsl getopt oniguruma5  
> poppler cairo gtk2 atk glib2 pango poppler-data Pantomime PRICE  
> TalkSoup netclasses Yap.app ImageMagick bzip2 a2ps psutils  
> gworkspace system-preferences PreferencePanes windowmaker
> Error: Status 1 encountered during processing.


You know, I am awfully motivated to use GNUstep but having to fight  
this hard just to install is too much!

Rather than focussing your efforts on changing the GUI with risk and  
no certainty of return, would you not be better off on focussing your  
energy into producing a nice self-contained .mpkg to install it  
easily on a Mac, and keep those updated? I think you would be MUCH  
better off. Do this, give it a bit of time and give the current look  
another chance - see if it catches on once people are actually given  
a chance to try GnuStep out.

You can do this with absolutely no risk of alienating anybody,  
probably with not too much work, and you are virtually guaranteed to  
win more users and more developers. Is this not a better business  
proposition than changing the theme?

This brings me on to the next question: who are you targeting? *You  
cannot appeal to everybody*. You try to appeal to everybody, and you  
will end up appealing to nobody. You have to pick your target  
audience. Is it Linux/KDE/Gnome developers or former NeXT/current Mac  
OS X developers? It looks to me like you're de facto targeting the  
former (all comparisons are with KDE/Gnome, all effort is going into  
having GnuStep bundled with Linux; but on a Mac, there isn't even a  
simple way to install the system).

I think this is a big mistake. There is a much greater barrier to  
entry coming from the KDE/Gnome world, and much more effort will be  
required to develop new applications from scratch. A much better goal  
for you is to persuade existing Cocoa/Apple developers (and mostly  
former NeXT developers, many of whom are now doing just that -  
developing for Cocoa) to port their apps to GnuStep. And to attract  
ANY of those, you absolutely MUST have a simple installer for GnuStep  
on Mac OS X. Moreover, trying to go after Linux developers has  
another disadvantage - those developers are less likely to understand  
what OpenStep is all about and are more likely to push it in the  
wrong direction. You can pander to their every demand, and GnuStep  
will never be good enough - until it's the same as KDE/Gnome.

Mark Grice wrote:

> If someone wants to keep the old style NEXTStep look and feel THAT  
> should be a branch. The main code has to move forward in this area  
> or no one will take GNUStep seriously.


I disagree. People look at substance, not form. We use computers to  
work. We ask: what can it do for me, not: does it look nice? I hate  
Apple's Aqua gummy, but that doesn't make me want to quit Apple.  
Apple's flakiness and lack of functionality does.

I think, on the contrary, it is GNUstep that should not take  
seriously people who decide against using GNUstep on the basis of its  
icons, and skins. Such developers are the sort who have made the mess  
of OpenStep at Apple; the value they will add to the project is  
negative.

In this spirit, rather than focus on the skin, what about some of my  
other suggestions:

> Strengths of GnuStep you can leverage:
>
> a) Objective C/OpenStep frameworks on non-Apple hardware (cheaper,  
> more robust/reliable than Apple);
> b) open source (could be important to users not wanting to be tied  
> to a proprietary solution);
>
> Forget - to start with - *everyone* wanting to use GnuStep, if they  
> want to. You need a niche that you cater for - to which you offer  
> something nobody else can, or, at least, to offer it 10x better  
> than anyone else.
>
> Niches you could potentially appeal to:
>
> 1) former NeXT users
> 2) technical / mathematical users (as NeXTStep)
> 3) Apple is not terribly solid and reliable. If you can be more  
> solid/reliable, you could potentially target users from the  
> "mission critical" market - like financial institutions.
>
> Here are some suggestions:
>
> #1: (i) more development frameworks. (ii) Reliable/bullet-proof/ 
> debugged frameworks. (iii) Faster/optimized frameworks.
>
> #2: applications. How about:
>      - to appeal to mathematical/technical market:
>        - a GnuStep clone of Mathematica notebook interface / do a  
> deal with Wolfram to develop one for GnuStep;
>        - GnuStepTeX
>        - ? etc. ?
>      - to appeal to the mission critical / finance market:
>        - an Objective C framework for derivatives pricing
>        - a Lotus Improv/Quantrix clone
>        - ? etc. ?
>      - educational software - to target schools
>
> #3: how about producing an install DVD which formats a PC's disk  
> and automatically installs a GnuStep/Linux distribution?
>      Or doing a deal with a PC manufacturer to sell - or for you to  
> sell - cheap PCs with GnuStep/Linux preinstalled?
>      Or doing a package deal to sell PCs with GnuStep/Linux pre- 
> installed cheaply to schools? Can you get a computer company to  
> donate equipment to schools and the GnuStep team installs the OS  
> and the GUI?