Re: [gnome-love] Linux Desktop Standard

[Martin Pool <mbp samba org>]

On 31 Jul 2003, Sridhar R <sridharinfinity yahoo com> wrote:

> There should be a standard in the OS community.  We
> will concentrate on End-Users side.  When somebody
> creates an application for a purpose, the application
> should try to satisfy that purpose (if possible - also
> related to that purpose) fully.  But in GNU/Linux
> world, I am seeing many softwares for a particular
> task.  For writing CDs, there are many (GUIs).  CLIs
> are very standard, but we are talking about the GUIs.

This is hardly unique to free software.  

On Windows or Mac OS, there are scores of (for example) FTP clients,
including those bundled with the OS, free software, freeware,
shareware, and purchased.  

There are many categories of applications where there are successful
replacements, even though they're shipped wit the OS.

> Again the presence of two different Desktop Managers
> is sometimes difficult.  I don't know how far is this
> issue for the users of GNU/Linux.  When you consider
> the Windows operating system, there is some standard. 
> For dialing up to the net, I need to use dial-up
> networking.  Microsoft doesn't provide more than one
> tool for a specific purpose.

Microsoft may provide only one but there is certainly more than one
tool *for the Windows platform*.  (And in some cases there is in fact
more than one MS tool for a job.)

Indeed from memory there are even alternative commercial dial-up
implementations for Windows, that have more features (scriptability,
etc) than the Microsoft version.

> Fine.  Now I am thinking to introduce a standard.

You are not the first person to suggest that...
 
> "Everything is one and the best".

OK, but who gets to decide which is "best"?  Reasonable people can
have different preferences for the complexity, featurefulness,
timeliness and licensing of the software they use.

In the software shipped with Windows, some Microsoft manager has made
an arbitrary (though researched) choice of what tradeoff is acceptable
for every component.  Users can accept it or look elsewhere, where
ISVs offer alternative components.

The situation is the same on Linux/GNOME, except that there is not a
single authority for the default install.  Instead, different
distributions can make decisions they believe support their customers.

If you are mainly concerned about distribution users seeing more than
one program for each type, then choose a distribution that does has
only one.  There are such distributions.  Perhaps you can start
SridharLinux, and ship exactly one of every application, and try to
make them all similar.  I think this is what many distributions are
tending towards today.

Telling individual developers that they must not develop a POP client
because others already exist is an unreasonable restriction of their
freedom, and anyway unlikely to work.  Who are you to tell them "you
must not write that program?"

The best you can say is "you would serve your customers better by
adding to an existing project."  But software doesn't work like that:
sometimes codebases are too old and hard to extend.  Sometimes there
is disagreement about stability/simplicity/features, or about
different approaches to the same end.

> But the final product (distro) is really mixed.

Which distro are you looking at?

> A desktop GNU/Linux system should consist of all required softwares
> each satisfying a specific purpose.  Being so, they should be as
> feauture rich as possible.  Whenever a developer write a software,
> he should confirm the fact, his or someone's project will be
> full-fledged end product for the respective purpose.

"As feature rich as possible" often comes at the cost of usability.
There is space in the world for both emacs and gedit.

> For a purpose, developers across the country

which country???

> should unite to work for it when they think about
> developing a software on their own for that purpose.
>
> I think there aren't many killer softwares available under Linux.
> If available, most of them are CLI or feature deficient.  There
> should be a standard for Desktop Linux.

There are some standards for how GNOME programs should behave

  http://developer.gnome.org/projects/gup/hig/
  http://freestandards.org/

etc

I'm not sure how you think that imposing standards will produce killer
applications.  The most innovative apps on all platforms have
typically gone beyond previous standards, and also have come from
upstart developers, not the OS-mandated standard programs.

> 1. gIDE has been a popular IDE for GNOME develpers
> once upon a time.  But now gIDE and the current most
> popular development tool Anjuta joined together to
> work on Anjuta Dev Studio.  
> http://anjuta.org
> This explains why this "One and the best" policy.

I agree that sometimes people fork or duplicate work needlessly.
Telling me about that doesn't help.  You need to persuade people who
are thinking of forking.  If you and all your programming friends
choose to contribute to existing programs rather than starting new
projects you will have made a positive step.

Here is an alternative example: sendmail is the standard mail daemon,
but many people have created alternatives to explore other approaches.
Many people consider Postfix or qmail much better than Sendmail.  It
would have been practically impossible to evolve Sendmail into qmail;
they are fundamentally different designs.

Should the authors of qmail and Postfix have been forbidden from
replacing sendmail.  (And how?)  In what way would the world have been
better off?

> 2. We have "powerarchiver" for windows.  No software
> in GNU/Linux is equivalent to this one.
> http://powerarchiver.com

What about file-roller?  If file-roller is missing features, now is
your big chance to improve it...

-- 
Martin