[OpenIndiana-discuss] I figured this deserved a separate thread

[Richard L. Hamilton]

On Nov 4, 2010, at 8:48 PM, Allan E. Registos wrote:

>> No, OpenIndiana both targets Desktops and Server 
> 
> 
> That is what most GNU/Linux distributions have done since the dawn of time. But the recent(well not quite recent) spat between Ingo Molnar and Con Kolivas (Linux kernel developers) does represent the major problem of Linux desktops : _responsiveness_. OI for desktop needs to be optimized for the desktop to IMO, unless there are technical issues. I don't know if I know what I am talking about since I do not know anything about systems programming, much less how the kernel works in the inside. However, all desktop users(non-devs) cares only of what happened in the desktop world ( playing games, watching videos, surfing, etc.) therefore, any OS must have some optimizations on this area 
> 
> 
> Cheers, 
> Allan

Solaris tries to be mostly self-tuning; knobs are for special cases,
needing them routinely in the long run has been considered wrong before now.

A desktop user would need their desktop stuff (video and HID drivers,
windowing system, GUI apps) to be more complete, reliable, and up-to-date
(pick two of those three, probably :-).

Other than that, a desktop user still needs decent disk and network performance,
and probably doesn't want files just up and disappearing either.

OTOH, they probably _don't_ care so much about dynamic reconfiguration,
self-healing, stable enough for year-long up times, etc.

But not all of the server features even take up memory (although they do
take up disk space) if they're not being used.  So getting rid of stuff
to turn a server distro info a desktop distro probably doesn't make that
much sense.  And even a desktop user could not possibly _object_ to
a system being _more_ stable than they need, as long as they don't
have to do anything extra or lose performance as a trade-off.

Likewise, a server probably has no good reason to have Bluetooth support;
it's just one more thing to go wrong that they don't need.  But if they
don't have the hardware, I don't expect the drivers would load anyway.


What I'm saying is that as long as package boundaries are reasonable
(for those that wish to minimize systems to get rid of much if not
_exactly_ all that they don't need), one single distro should work
find for servers and desktops alike.  

Those with interests particular to one or the other should participate
or otherwise support those activities working on parts relevant to their
interests.

Of course as you move into laptops and even smaller mobile devices,
the specialized requirements increase, as does the need to squeeze out
all the parts not needed, to be able to fit.  Less secondary storage
(disk or flash), less RAM, using power efficiency increasingly critical,
more specialized hardware and even GUI (for smaller than laptop) elements,
etc.

But down to a laptop should IMO still work fine as part of the same
distro that runs on servers.  Below laptop, minimization becomes critical,
as do alternative inputs like touchscreens or voice including both driver
and UI support; and the whole packaging and maintenance structure needs
to be more appliance-like and less like a general purpose device.

So unless someone is planning on running a mainstream CPU (for Solaris read
SPARC or x86) version of Solaris or OpenIndiana on a tablet or smaller
device, I don't see the need for distros specialized to the type of platform.

One more thing: I'm not saying it's impossible to come up with desktop
specific optimizations.  I _am_ saying that if such optimizations become
pervasive throughout the OS without being isolated and modular to the
poing of being a seen via a single installation-time knob, then staying
compatible with the base OS (commercial Solaris) becomes increasingly
difficult, and which point more than just control over some of the code
is forking: community, expectations, resources, and eventually API or
internal interfaces would start to diverge too.

Forking makes sense about as much as carving countries up into smaller
countries - only if the whole is dictatorial, grossly inefficient, or
is neglecting or exploiting some regions for the sake of others.

Developers and users support each other.  Developers...develop.
Users pay, buy beer, or at least test and provide constructive feedback.
(a user that does _none_ of those, even when the latter is made
reasonably easy, is a parasite, even if the license is free enough to
allow them to be a parasite)

As long as a distro doesn't prevent desktop or server developers and
their respective users from working together, nor prevent the developers
across those areas from working together, one distro should be good enough.