Philip Callan wrote:
> Alan Hargreaves - Product Technical Support (APAC) wrote:
> Well, they /were/ interested in all those nice drivers, but lost
> interest rapidly when they realized what it would mean, linking to the
> kernel.
The whole point is that you don't have to have the kernel available to
build a module for Solaris. A good example is the network drivers that
Muryama-san provides from his website. Those drivers build completely in
the absence of the Solaris source tree and then simply drop in.
What we were intersted in was not distributing those drivers ourselves,
as we obviously were precluded from doing so, rather distributing a kit
that would allow end users to do the port themselves (distribution then
would not enter the equation).
> AIUI, Sun isn't precluded from using LGPL drivers are they?
I believe LGPL is ok. It's RMS' interpretation of linkning that kills us
distributing even distinct loadable modules that are licenced purely
under the GPL.
>> There are other avenues to getting driver support other than the only
>> one that you suggest.
>> For example:
>> - request that the author provide the code under a multiple license
>> (been done in a number of instances).
> What happens when the code is a work of multiple authors?
More tricky. Now I'm not in the area that does this negotiation, but I
imagine that you would have to start with the initial author and work
down the tree. I guess after a bit it becomes a story about diminishing
returns.
>> - (shock, horror) write it ourselves. Over the last year, there have
>> been a huge number of engineers hired with precisely this aim in
>> mind, and there has been a lot of driver support added for much
>> current hardware.
> Good, nice to see some of the cash MS paid you folks going to coders.
These folk were hired way before that cash came. Something else that a
lot of folk do not realise is that even though we publicly "deferred"
Solaris 9, engineering did not stop. Indeed, a lot of the engineers were
running Solaris 9 on their notebooks and initial development actually
happened there! Even without that, there was an unwritten rule that we
not break the x86 builds with anything that we were doing. Solaris on
x86 still had to build and boot, lint free, warning free.
> Like I said, Sun's support for non-Sun or Sun partner HW was one of
> their weakest points, on x86 at least. Otherwise 'standard' devices,
> NIC's and SCSI controllers, were finicky about what was/wasnt supported.
It is probably worth commenting at this point that when I bought my off
the shelf Dell Inspiron 8500 notebook, Solaris installed and ran without
modification. I was short only two drivers.
1. The broadcom ethernet that was in it (thank you Muryama)
2. The broadcom wireless (which we still don't have the specs for - nor
does the Solaris or *BSD* communities).
> Now for all I know, this could be because of vendors deviating from some
> standard, or protocol, and Sun not keeping up with the times, but if the
> OSS community can design drivers for all these devices, there is no
> reason paid engineer's shouldn't be able too.
This is generally true. Unfortunately there are vendors out there who
don't play nice with specifications. If we have the specifications,
drivers can be written.
>> - speak to the hardware vendor about porting. This is also taking place
>> with a numer of hardware vendors.
> Which is definitely an area where Solaris / Sun have an advantage, Sun
> can probably bring some more effective pressure on x86 HW vendors to get
> them to support non-Windows platforms, although given their agreement,
> I'm more inclined to think they will try to convince HW manufacturers
> not to support those 'Other' non-Windows x86 flavors (BSD/Linux) as with
> Sun/Windows, they can still get away with distributing binaries, and no
> source.
Just because Sun is a big corporation does not immediately mean we are the
bad guys. I think we have better things to do with our time than to try
to convince vendors not to place nice with other *ix variants.
> I figured as much when they started drawing a distinction between 'Open
> Solaris 10' and the Solaris 10 'with *encumbered and proprietary bits*'
Open Solaris 10 - no such beast.
The open source product is/will be called "Open Solaris". Solaris 10 is
the Sun Branded product that is released some time in the next month or so.
Now we are starting (unfortunately) to enter an area which is still
under NDA. What I *will* say (however) is that we are looking to open as
much as we possibly can. The issue of drivers that we do not own the
source rights to will be dealt with in such a way that Open Solaris
*will* be fully buildable.
> You have to have something to offer as incentive to get them on the
> hook, if the free version had ALL the functionality of the paid version,
> people would have no reason to get the 'real deal'
The big thing that one would get with the "real deal" is a system that
can be put under Sun Supprort and get the full support that we offer,
including such things as round the clock work on bug fixes.
> Support I assume can be contracted from Sun regardless if the client
> uses the Open Solaris product, or the Closed, with perhaps a difference
> in price, discounted for using the 'official' release, with 'approved'
> software packages you have tested.
Support of Open Solaris is again something that I don't think I can go
in to (yet).
Really, we are not the bogey man that many folk paint us as.
alan.
--
Alan Hargreaves - http://blogs.sun.com/tpenta
Kernel/VOSJEC/Performance Engineer
Product Technical Support (APAC)
Sun Microsystems
