[OpenIndiana-discuss] [illumos-Discuss] branding for illumos/openindiana

[Richard L. Hamilton]

On Jun 22, 2011, at 9:14 PM, Gregory Youngblood wrote:

> 
> On Jun 22, 2011, at 5:48 PM, Richard L. Hamilton wrote:
> 
>>> The biggest problem I have had with people trying OpenIndiana (and yes, Solaris too) has been hardware compatibility. One colleague summarized it this way: OpenIndiana, where Linux was in 95. He's referring specifically to drivers for various things, especially controllers. He went through 3 or 4 cheap controllers between NewEgg and Fry's Electronics, even buying one from NewEgg specifically since it was supposed to be compatible with OpenIndiana, and none of them worked. He never got a finished installation to play with. [The problem was the vendor revved the board, slight change in chipset, but didn't change the SKU so NewEgg never noticed the change. It only affected OI, the board continued to work just fine in Linux.] 
> ....
>> 
>> Funny, I've bought a SAS controller (LSI) and a Gbit Ethernet board
>> (Intel) that both worked just fine on _SPARC_ even, and the Ethernet
>> even worked on Solaris 9 (although I did have to scrounge for an
>> open source driver there, since the e1000g driver wasn't in Solaris 9).
>> (I don't know if I could boot over the net on the GbE since the
>> card has no fCode, but I don't really care, since if I needed to
>> that badly, the old interfaces are still in place.)
>> 
>> Point being that if one is careful enough about compatibility (to
>> include sticking with brands that don't randomly juggle chipsets on
>> otherwise unchanged motherboard part numbers), and checks the HCL too,
>> one can do ok.  Just not on whatever random hardware one happens to
>> have.
> 
> That's just it - at least one of the cards he purchased in an effort to set up an OpenSolaris or OpenIndiana file server came directly off the HCL, brand/chipset/etc. The other cards were too. 
> 
> This was a gentleman that had at least 7 years working with server hardware and probably over 15 years working with computer hardware in general, in other words he was not a n00b. He used the HCL when selecting the hardware he tried to use. He followed reports about specific cards with chipsets that supposedly worked, and ultimately also tried purchasing the exact brand and model in the HCL itself. Despite doing all the right things, he didn't end up with a functional system, and after several days got fed up and moved on. 
> 
> So the glib answer of being "careful enough about compatibility" and dismissing people's bad experience is part of the problem that has to be overcome.
> 
> I will admit the extent of his experience was atypical, I've been able to get Solaris, OpenSolaris, and OpenIndiana to work on just about every computer I've tried with only one real exception (a netbook the DDU reported 0 driver problems in the live image, but the install didn't have working networking (wired or wireless) and some other faults).
> 
> Like he said, [OpenSolaris/OpenIndiana] - where Linux was at in 1995. And based on his experience, he's right. He did what _should_have_ worked and it didn't.

Pending the magic availability of infinite drivers, the simple answer
would be a better HCL.  As in, track those brands that change
chipsets without changing part numbers, and whenever something
_doesn't_ work that the HCL said did work, encourage people to report
that too, preferably with any available distinguishing information
(revision number or whatever).

I think the basic problem is that x86 motherboard manufacturers,
except those that are specifically targeting high stability
uses (servers, for instance), assume that if they've made sure
it runs the latest Windows, their job is done, and nobody _other_
than the board manufacturers can possibly keep track of
each new board.  No _one_ person, or even small team (lacking
a budget to buy samples of everything regularly) can.  But an
aggregate of reports would sure help.

There's also the "Oracle Device Detection Tool".

Or if one needed something of one's own, one could roll a bootable
DVD or thumb drive distro, that hopefully stores might allow one
to try, as a compatibility test for the hardware.  Doesn't do any
good for build-it-yourself or mailorder of course, but those are
pretty much caveat emptor anyway.  At least it's more tools...