Does FreeBSD 6.0 fully support PCI-Express?

[Ted Mittelstaedt]

>-----Original Message-----
>From: Micah [mailto:micahjon at ywave.com]
>Sent: Thursday, November 24, 2005 8:11 AM
>To: Ted Mittelstaedt
>Cc: Hans Nieser; freebsd-questions at freebsd.org
>Subject: Re: Does FreeBSD 6.0 fully support PCI-Express?
>
>
>>
>> Micah, if this has changed, please cite where.  I myself also
>happen to
>> have a system with an onboard nvidia card so I really am interested,
>> not just trying to flame-bait.
>
>I think I understand your claim.  Source code with an open source
>license is not Open Source unless it is actively maintained by someone
>and has freely available specs.

No, not exactly.  This is a special case with device drivers.

A simple standalone program that does not interface with hardware, if
it's source is open with an open source license, it's open source,
even if it is not actively maintained, and even if nobody has published
a flowchart or logic diagram that indicates how the code works.

Such a program can be modified or maintained by anyone, if they are
competent enough.

But device drivers are different cases, because even if the source
is open, and licensed as open, unless you are able to determine
how the hardware works from looking at the source code, it really
isn't open source because nobody can modify it.  Nobody that is,
except someone who has the hardware technical documentation.  And
that documentation is not something Nvidia gives out, even under
NDA.

The author of NV, Mark, is an Nvidia employee, so he has access
to this data and can modify the driver.  But nobody else can
modify the driver who doesen't work for Nvidia.  Thus it's
immaterial - for a device driver - if source is open or not or
source is licensed as open source or not, because other people
are prevented from working on the driver.

It would be like if I patented a software algorithim and released
"open source" for it, then started suing everyone for patent
violations who simply used that source. (NOT copyright violations)
Kind of like Unisys and the .gif file format.

>Under that criteria, I guess NV isn't
>open source.
>

There's an open source organization which is trying to establish
branding on the name "open source" who has a bunch of criterian
that they claim a program license must meet to be termed "open source"
I don't hold with that, but if you do I do not think that the source
for the nv driver meets their criterian either.

I view a program as being open source based on what the copyright
holder intends with it.  For example, I don't have a problem with
a copyright holder claiming copyright on a program then writing a license
that only permits people to download the source and compile it
and use it in their own projects, or for custom projects they
are doing for other people, but prohibits people from compiling
binaries of the program then selling those binaries as standalone
programs, or using the source in software that they are selling
standalone binaries of.  To me, that is "open source" but to
a lot of people it isn't.  To GNU it isn't, but rather than
writing a license that bars selling software, they wrote a
license that requires source to be provided if you do sell the
software, in the hopes that this would kill enthusiasm among
people for selling open source.  Goals of GNU and the hypothetical
copyright holder are the same, but approach to that goal is
different - thus GNU claims it's own stuff is open source and
the hypothetical copyright holders stuff isn't.

Ted