2012/11/12 Jonas Bonn <[email protected]>:
> On Mon, 2012-11-12 at 06:06 +0200, Stefan Kristiansson wrote:
>> On Sun, Nov 11, 2012 at 10:16:25PM +0100, Franck Jullien wrote:
>> > As I told you when we met, the tiny SPI core is a good example.
>> > This core as a driver in the Linux mainline (spi-oc-tiny.c). However,
>> > on the opencores website, it doesn't have a OCCP stamp. This
>> > should not be possible....Plus, the official maintainer should be
>> > opencores because the core has to remain very stable. Any change
>> > in the core should be acked by opencores gurus and checked against
>> > the linux driver.
>> >
>>
>> Although I agree with most of the other things you brought up,
>> this example is good.
>> Who are those opencores gurus that should maintain it?
>> I for ones have never used that core and probably never will,
>> so I don't really care much about it.
>> It happens to have a driver in Linux mainline since the
>> creator of the core submitted a driver that he had written.
>
> As I recall it, the Linux driver was submitted by one of the NIOS folks
> and wasn't the author/maintainer of core himself.
>
> In my opinion, OC would better serve as a ward of _quality_
> specifications than as another sourceforge/github.  As long as there's a
> single, complete, and correct spec hosted by a reputable central
> organization, then it subsequently matters little how many
> implementations of the core/driver there are, where they are, or who
> maintains them.  The spec's, today, I find a bit too "unfinished"... and
> perhaps a bit too fluid; if there's anything a "guru" should be watching
> for it's incompatible changes to the specification.
>

I used the wrong term when said "guru". Should have said opencores
community. That's why we sould have an opencores mailing list and
changes be submitted to the list for review.

> I worry a bit that the 'compatible' tag on the Linux drivers is so
> tightly coupled to implementation today, whereas as it really should
> just be coupled to the OC specification revision and, thus, be _mostly_
> implementation independent.
>
> So my suggestion would be:
>
> i)   make spec management central and do quality control of this
> ii)  when searching for a project, it leads you to the spec and,
> secondarily, points you to a list of available implementations (and
> available drivers, if applicable)
> iii) "certified" stamps on the site go on the implementations, not the
> project as a whole...
>

That's indeed a very good idea. It would solve a lot a problem.

>> If that guy isn't actively maintaining that project anymore,
>> and you want to step up and give it some love, that's great.
>> If he's not interested in being the maintainer anymore,
>> then he'll probably gracefully handover the maintainership
>> to you, just like it would happen in any other open source project.
>
> Just fork the project and do a better job maintaining it than the
> original guy and it's yours... if you cannot reach the original
> maintainer anyway, then this isn't even considered unethical.  But, as
> always, be careful what you wish for...!
>
> /Jonas
>
_______________________________________________
OpenRISC mailing list
[email protected]
http://lists.openrisc.net/listinfo/openrisc

Reply via email to