[oi-dev] Resignation as OI Lead

Thu Aug 30 20:18:30 UTC 2012

On Aug 30, 2012, at 12:20 PM, Jose-Marcio Martins da Cruz <jose.marcio.mc at gmail.com> wrote:

> 
> Dear Garett,
> 
> I think I strongly disagree with you.

I think you misunderstood me.  More below. :-)
> 
> garrett.damore at dey-sys.com wrote:
>> Dear Alasdair,
> 
>> 
>> The model of OpenSolaris is broken. The model of OpenIndiana following
>> OpenSolaris is broken. The illumos model is following the successful Linux
>> model. This is exemplified by distributions such as the commercially supported,
>> general purpose OmniOS, Joyent’s SmartOS, Delphix OS, and others.
> 
> In our context, education/research community, we expect an OS coming from some kind of model, a model like Linux Debian or FreeBSD, with a free OS (no strings attached to it), but with people doing business around it with some kind of service (support, installations, solutions, ...). The kind of model we had with Sun was also fine for us : in the 80's and 90's with had a site license for some number of nodes. After that, till the moment Sun was bought by Oracle, the OS was free, but we paid for support. This was also OK.
> 
> Openindiana is the Ilummos distribution which best fits what we expect.
> 
> IMHO, it's an error making Illumos follow the Linux model for the simple reason that the user base size isn't comparable. Still IMHO, because of the difference in user base size and the amount of developpers, it should be better to aggregate resources in order to have, for the moment, a solid distribution instead of all these distributions (OmniOS, Joyent's SmartOS, Delphix OS, ...). Maybe this is what is killing Openindiana. You should think about.

Actually, when I tried this, the result was illumian, which didn't work out so well.

All of the distributions you list above are being developed by *commercial* entities that have their own business needs.  We collaborate around a common kernel, and there may be areas where there is some collaboration with other upstreams, but the distributions are different because they have different *purposes*. 

The presence of these competitors is most definitely *not* what is killing OpenIndiana.  (Although, I'm told that some parties have switched from OI to SmartOS.  But I think that underscores the real problem with OpenIndiana, which I'll get to shortly.)

> 
> Before the message of Alasdair, I was just preparing some dozen of servers to get into production in our organization, running some infrastructure applications (DNS, mail servers, directory servers, NFS servers, a lot of web servers, ...), based on Openindiana.
> 
> After your message, I looked for the distributions you mentioned above. None of them fits the model I want.

Well, I'm not sure what the model you want is.  Of course some of those are commercial (all of 'em really, but then so are most Linux distros).  SmartOS, OmniOS, illumian, and OpenIndiana should all be reasonable technology bases for what you are describing above, and they are all basically open source.  I think you should look at the alternatives -- but then again if OpenIndiana works for you, great!

> One of them even doesn't have, in their site, a download link, but have a price page. So, we're coming back to some kind of closed model, the same Oracle model all of you are criticizing.

> At another one, it seems that some features are disabled if you don't have a support contract (zfs send/receive). So, again, back to the Oracle closed model. If I shall go back to a closed model, maybe I'll prefer remain at Oracle.

Again, you misunderstood what I was talking about.   I wasn't talking about open vs. closed at all.

So let me clarify:

What is *broken* was the model of slavishly trying to follow OpenSolaris, or to be an "open & free" alternative to Solaris 11, servicing servers, desktops, laptops, and both SPARC and x86.  That was the model that OI started with -- to simply package up the bits that Oracle was providing, try to match it to an illumos kernel, and package the whole thing up. 

What's broken about this is two fold.

First, from a technical level, trying to retain and use packages from an upstream like Oracle, where there are dependencies upon closed bits, and flags days and interface boundaries where we we only get half of the changes, is untenable in the long run.  It's been doomed to failure since inception.

Second, and probably more significantly, the *vision* is busted.  OI had *no* vision except to follow Oracle's lead.  Even Oracle abandoned OpenSolaris and the desktop, but OI tries to muddle on with no clear "vision" about what sets it apart.  There is no "innovation" in OI, really.  Too many people want too many things from it (server, desktop, compatibility, SPARC vs. x86), to the point that it can never really take the necessary steps to excel at any one thing because doing so might make it worse at another.  OI became jack-of-all-trades, master of none.

(In fact, I'd argue that the desktop focus has been a huge drag on OI -- keeping X, Gnome, and all the various other components updated is such a huge job that its almost impossible to do it as a hobby.  And as far as I know, the only commercial company to have done any significant investment in that for OI is EveryCity itself - and that is now a thing of he past.  Ditching the desktop stuff would surely piss off a fair number of people, but from what I can see none of those people are actually providing the kind of investment -- in either money or technical support -- to make the sustaining of a desktop illumos distro a viable effort.  Your own comments about wanting a "free" (as in beer) option underscore that for me. Nexenta has some direct experience with this -- their first release was a desktop distro.  Which was a huge amount of work, and completely commercially unviable.  Even Oracle with their not-insignificant resources has realized the futility of trying to maintain Solaris on the desktop -- they've abandoned that market segment entirely.)

Now, the other distributions have much clearer purposes.

* DelphixOS drives their database enhancement technologies.
* SmartOS is designed to be the finest hypervisor technology for data center virtualization on the planet
* OmniOS is a designed to run server software (exactly the sort of stuff you listed above, actually)

And of course, at DEY we are working on a variant designed to server another special purpose -- one that is about as far from a general purpose computing platform as you can be.

These are all leaders, innovating in some way or another.  Can anybody point to an area where OI has contributed any significant innovation to the ecosystem?  I can't.  OI is a follower.  It filled (and perhaps still fills) an important role for the community, but lets not pretend that OI is the seat of excellence or innovation.  Other distros fill that role.

> 
> Now that I know what you really think about the future of Illumos and their distributions, I may definitely consider another OS : Linux of FreeBSD.

I don't think you really understood what I think.  However, you should consider what it is you really need.  Why did you choose OI in the first place?  Maybe OmniOS looks pretty good -- unless you need a desktop distro.  In which case, yeah, OI is about the only game in town for illumos. 

In fact, last year I stopped using OI on my desktop.  Why?  I got tired of fighting the constant battles to deal with application integration.  I need support for things like GotoMeeting, Skype, and other software that my customers and colleagues use.   OI can't get there -- there isn't enough market share to make it interesting to those desktop ISVs.  So, now I use a Mac, and my life got a lot better.  I can focus more time on engineering (for illumos!) and less time worrying about printer drivers, and audio/video problems, suspend compatibility, etc.   Of course, I would *never* consider using a Mac as a *server* (at least not running MacOS).  I don't trust HFS+ further than I could throw it.  So illumos as a server makes a lot more sense for me.  But I'm *personally* done fighting the desktop battle.  Both MacOS X and illumos have a lot of excellence in them -- they just differ very very much in what things they are excellent at (right tool for the job, etc.)

Does that mean other people aren't still fighting that battle?  Of course some are.  (See Martin's recent postings about X and SPARC.)  So even if *I* think that battle is lost, others are free to continue.  That's the great thing about a community.

Heck, I'd welcome an effort to create a truly great desktop on illumos.  But if such a distro arises, I hope that they will make that a *focus*, and worry about doing that the best way they can, rather than also trying to be a great server, virtualization, and so forth.  I'd like to see some innovation here - not just another compile of X of illumos.  Show us how illumos can be compelling here - a superior solution on the desktop or laptop to Linux, BSD, Windows or Mac, if you're going to take that effort.

And the great thing about diversity in distributions, is that each is free to pursue a course that lets them focus on being great at what is important to them.

*That* is the model I really want to see -- the pursuit of excellence. 

And I think the example distros I listed before are demonstrative in this regard.

	- Garrett