[linux-dvb] RE : [RFC] Let the future decide between the two.
P. van Gaans
w3ird_n3rd at gmx.net
Thu Sep 25 18:40:54 CEST 2008
On 09/25/2008 02:10 PM, Thierry Lelegard wrote:
>> De : linux-dvb-bounces at linuxtv.org
>> [mailto:linux-dvb-bounces at linuxtv.org] De la part de Janne Grunau
>> Envoyé : jeudi 25 septembre 2008 12:48
>> À : linux-dvb at linuxtv.org
>> Objet : Re: [linux-dvb] [RFC] Let the future decide between the two.
>> On Thursday 25 September 2008 08:45:28 Michel Verbraak wrote:
>>> I would like to propose the following:
>>> - Keep the two different DVB API sets next to one another. Both
>>> having a space on Linuxtv.org to explain their knowledge and how to
>>> use them. - Each with their own respective maintainers to get stuff
>>> into the kernel. I mean V4L had two versions.
>>> - Let driver developers decide which API they will follow. Or even
>>> develop for both.
>>> - Let application developers decide which API they will support.
>>> - Let distribution packagers decide which API they will have
>>> activated by default in their distribution.
>>> - Let the end users decide which one will be used most. (Probably
>>> they will decide on: Is my hardware supported or not).
>>> - If democracy is that strong one of them will win or maybey the two
>>> will get merged and we, the end users, get best of both worlds.
>>> As the subject says: This is a Request For Comment.
>> This is complete nonsense, distrobution packagers shouldn't
>> decide which
>> API should be used, the API and all drivers should be in the kernel.
>> Having two tree is at best fragmentation and at worst a whole lot of
>> duplicated work.
> Having the two coexisting API is a COMPLETE SOFTWARE DESIGN NONSENSE.
> This would not be a "cathedral to bazaar" transition, as someone wrote
> on the subject, it would be a "bazaat to mess" transition.
> I have no technical opinion on Multiproto vs. S2API since I have been
> using only DVT-T device for the last two years. But I have more than
> 20 years of experience in software design and that would be for sure
> the worst decision ever.
> On a user point of view, Janne's point is the most important one:
>> That should a user do if he has two devices which are only
>> supported by one of the trees? That's bad luck?
> And this is not only a Multiproto vs. S2API issue. As I mentioned,
> I use only DVT-T devices. I have 4 of them, all working on Linux
> for months or years and there is no one single repository supporting
> all of them at the same time. Most are supported by
> http://linuxtv.org/hg/v4l-dvb, another one needs
> http://linuxtv.org/hg/~anttip/af9015 plus other patches.
> And this is not a transitional situation, it lasts for months.
> This is an endemic but unacceptable situation. And, again, this is
> not only Multiproto vs. S2API. IMHO, linux DVB has a real leadership
> problem. There are just too many different forks which could be accepted
> during transition periods (development and validation of a driver)
> but which cannot survive that long.
> All these various trees contains technically good code. So, this
> is not a technical problem. Failing to merge them is a leadership
> problem ("cathedral companies" would say a "management problem").
> Imagine what would be the kernel today with that kind of methods?
> But for the kernel, there is Linus, a leader that you may like or
> not but that everyone respect (and respect the decisions of). The
> situation in Linux DVB is not like that, unfortunately.
> Since Linux DVB is a subsystem of the kernel and since there is
> no undisputed leader in Linux DVB, why not asking for Linus'
> arbitration? He started to get involved AFAIK.
> Currently, all arguments are like:
> - "This API is technically better" (not a technical problem we told you)
> - "Give me technical reasons" (same)
> - "This was voted" (validity of this vote is disputed -> leadership pb)
> - "You don't like me" (personal problems)
> One way to get out of this loop is to call for Linus.
> linux-dvb mailing list
> linux-dvb at linuxtv.org
You know, I've been thinking about setting up a new project. It's not
going to happen because I'm not a programmer and it would get very
messy, but my idea was to fork v4l-dvb and make it support as much as
possible. The AF9015, TT S2-3200, TT S2-3600. There is code but all in
seperate repositories and patches. And those three are not all (AF9015
is apparently now merged).
I still can't help thinking that if I wouldn't have asked why the AF9005
driver a while ago wasn't in hg yet it STILL wouldn't be in hg.
At the dev front, there should be some people with the desire to support
as many devices as possible (should already be the case), actively
asking driver developers how the're doing and if their code is ready to
be merged (or maybe a seperate team is required for such a thing?). And
possibly help if required. And if there is a driver that's working fine
_but_ it needs some code cleanup or something else minor like that, it's
NOT a reason to not include it at all!
I think there are even several driver developers who don't know they
have to send a pull request but assume this happens automagically when
it's known on the list that there is a working driver.
I just don't understand what the problem is. Is money required to buy
the devices to test them properly? Should a team be set up to
communicate with driver developers? Those shouldn't be unsolveable problems.
And what has happened with Multiproto/S2API.. I hardly understand. Maybe
we should have asked for Linus a long time ago. Have him take a look at
everything, let him make a decision and continue his chosen path. Maybe
not as democratic but it would be a whole lot faster, I think.
P. van Gaans
More information about the linux-dvb