[oi-dev] [developer] GSoC? Decision time....

[Nick Zivkovic]

I'm replying to _everyone_ not just Kieth. It seems this discussion
spans several lists...

Just to throw in some of my personal experience.

While a freshman, I had only one computer at my disposal. Running
OpenSolaris (later OpenIndiana) was very painful. I couldn't NOT run
it because it was the only OS with ZFS, DTrace, and Zones. From the
perspective of having only one machine, Illumos is a mixed blessing.
It cripples you by not allowing you to easily do desktop-ish things,
but gives you an __amazing__ platform for (especially C) programming
and data storage. I think that the Illumos-desktop folks see the
_potential_ of what Illumos _could_ be on the desktop and want to make
it happen.

Long story short, I changed my strategy from zealously trying to turn
OI into _the one true OS_,  into a combined-systems approach: I have a
SmartOS server that holds all of my data, zones, and KVMs, and that I
use for >90% of my programming (over SSH). I have three client
systems: a Mac laptop for normal consumer-things, a Thinkpad running
FreeBSD for client-side DTracing of desktop apps and hooking up to my
server, and an outdated Ubuntu tower for anything the other systems
don't cover. I often code on SmartOS and then deploy on FreeBSD and my
Mac (with minor changes).

Even if the Illumos-Desktop effort reaches application-parity with
Linux, there is still the major problem of device drivers. Linux (even
FreeBSD) has a huge advantage here, that is difficult to match without
a large supply of enthusiastic device-driver writers that all have
diverse hardware. OI hasn't been able to recognize my WiFi card for
years, and SmartOS can't even boot --- it's not exactly level playing
field.

Having multiple, specialized systems while not _physically_
economical, is very economical in terms of _time_ --- you don't have
to waste it reinventing so many wheels.

Of course this approach is broken over bad networks --- which sadly
are too common in parts of the world.

That said, a desktop variant of Illumos could leverage the clearly
innovative and high-quality kernel tech in some cool ways, but I can't
see this making a dent in the market share of Linux --- let alone that
of Mac, iOS, Windows, or Android.

Desktop-Illumos (OpenSolaris) was useful around 2006ish as a way to
get young people (I was still in highschool) to use and understand the
benefits of DTrace, ZFS, Zones, etc. It was a great gateway. I think
that the Illumos project doesn't need a desktop-variant per-se. It
needs a _gateway_ through which it can recruit the next generation of
systems programmers for the project. Seeing how the trend is that
tablets are devouring desktops and laptops for most of the population,
I think Illumos needs a new "gateway strategy" through which it can
easily recruit young blood. Just as OpenSolaris (the distro) was able
to frame the benefits of its unique technologies in the desktop
context, I think that Illumos needs to find a way to frame the
benefits of its unique technologies in the context of whatever
16-year-old computer-enthusiasts use these days to get things done /
have fun / etc.

Anyway, I'm _not_ convinced that GSoC is such a gateway...

Cheers,
-Nick

On 2/14/14, Keith Wesolowski <keith.wesolowski at joyent.com> wrote:
> On Thu, Feb 13, 2014 at 11:52:53PM -0500, Gordon Ross wrote:
>
>> I'm not sure if you were aware of this, but the GSoC program
>> organizers have in the past expressed a prefer that related projects
>> "band together" with shared applications.  That makes less work for
>> them.
>>
>> I recommend we tell GSoC in our application that we will entertain
>> projects using any and all illumos distributions.
>> OpenIndiana is one of several good choices.  I see no harm in
>> mentioning their inclusion.
>> Would you be more comfortable if we mentioned other open-source distros as
>> well?
>
> I'll turn that on its head: would everyone else be comfortable with the
> addition of SmartOS-specific work to the projects list?  I certainly
> would not be, if I were a partisan of a different distribution.  And I
> don't even consider most of the projects the OI folks have listed as
> being in any way related to the development of an operating system (nor
> would many SmartOS-specific projects I might come up with).
>
> But I guess that's just me.  Bluntly, leading with OI or treating it as
> special is thoroughly and uncompromisingly insane.  Pretending that, in
> 2014 no less, achieving desktop application parity with 1997 GNU/Linux
> is an important goal in operating systems development probably covers
> half the syndromes described by the DSM-5.  In the community I *thought*
> we were a part of, that's a nonissue, because the insanity is kept off
> to the side, separate from the core effort and merely one consumer among
> equals.  To each his own.  Live and let live.  What I'm learning here is
> that many people don't see things that way, and want to explicitly
> define illumos and OI as having a single, shared vision.  If that's
> what's happening here, I'll be aboard the next train out of this
> nuthouse.
>
>
> -------------------------------------------
> illumos-developer
> Archives: https://www.listbox.com/member/archive/182179/=now
> RSS Feed:
> https://www.listbox.com/member/archive/rss/182179/21175392-65b93de2
> Modify Your Subscription:
> https://www.listbox.com/member/?member_id=21175392&id_secret=21175392-568c67de
> Powered by Listbox: http://www.listbox.com
>