Re: [linux-dvb] Multiproto API/Driver Update

[Julian Scheel]

Michael,

On Monday 15 September 2008 17:42:06 Michael Krufky wrote:
> In summary, the bottom line is this:
>
> Manu did a great job with the multiproto API, people were happy using
> it, and all of the LinuxDVB developer community was happy with the
> work that was done, and we all wanted to merge it ~ two years ago.
>
> Back then, Manu said that it wasnt ready, so for some time we waited
> for him in hopes that he would agree that it was ready for merge.
>
> As more months went by, Manu was asked if he would be merging his
> changes, and he kept answering to the effect of "it's not ready yet,
> but should be in a few weeks"
>
> Months and months and months went by since then, with an occasional
> ping to Manu, with the reply "not ready yet" ...
>
> Two years is a very long time to wait for a new API, especially
> considering that it was functional from the start.  It was looking
> like Manu may not have had any intention at all to merge his work into
> the master v4l/dvb development repository --  It should be not be
> surprising at all that Steven Toth felt the need to come up with his
> own solution.
>
> Steven posted a proposal for his API expansion solution, and he
> received positive feedback.  Immediately, Manu broke out of his
> silence and sent in a pull request.
>
>
> Is there malice here??  No.  There is development, and developers
> looking to move forward.  Nobody is at fault.
>
>
> I have posted the above just to clarify what the "politics" actually
> are, here.  The only real politics going around are those that are
> accusing others of politics themselves.
>
> Had Manu been willing to merge his work earlier, this would have all
> been a non-issue.  However, now there is an alternative proposal on
> the table, which appears to be more robust and may have more potential
> that the multiproto proposal.
>
> Does that mean multiproto is disqualified?   Absolutely not.
>
> Does the fact that multiproto was there first mean that it will be
> merged without question now that it is suddenly available?  No, not
> necessarily.
>
> What does it mean?  It means that now there are two proposals on the
> table.  Two ways to solve a problem using different ideas and methods.
>
> The end users that have piped into the discussion are mostly concerned
> with which API demonstration repository already contains support for
> their device.  This should have no bearing whatsoever on the decision
> of the linuxDVB API extension.  All drivers will be ported to
> whichever solution is decided upon.
>
> Now is the time to examine these solutions from a developer point of
> view, in terms of how this affects kernel developers and application
> developers alike.  There is no reason to rush into things just because
> a pull request has been made -- the outcome of this decision is highly
> important, and we will have to live with the decision for a good long
> time.

Thanks for your version of the history. I just have to say I can't really 
agree with the way you describe the history. To point this out I looked up 
some of the old threads...

So everything started in 2005 with initial proposals for a DVB-S2 extension of 
the API by Marcel. In early 2006 there were some discussions about it on the 
lists:
http://thread.gmane.org/gmane.linux.drivers.dvb/23914/focus=24030
http://thread.gmane.org/gmane.linux.drivers.dvb/24239

At that thought not much (if any) capable hardware was available, so the idea 
was put off for the moment.
Then in April 2006 Manu started to work at the things and provided a first 
draft based on the changes from Marcel:
http://thread.gmane.org/gmane.linux.drivers.dvb/25401

An initial driver for KNC cards was provided by Manu based on this API 
proposal. After some discussions on 05 May 2006 Manu requested for a pull of 
the API:
http://www.linuxtv.org/pipermail/v4l-dvb-maintainer/2006-May/001006.html

Immediately followed by Johannes stating that he is not satisfied with the API 
yet and suggested a rework:
http://www.linuxtv.org/pipermail/v4l-dvb-maintainer/2006-May/001007.html

At that time rework began while in parallel some people (including jusst 
technologies) started testing the first drivers. As expected they were still 
far away from running perfect.

So it took a while to come to obvious progress. In January 2007 Manu announced 
the mp-stb0899-c5 tree - not even the current multiproto tree - which included 
the results of the rework. Some testing was done on that by more people.
http://thread.gmane.org/gmane.linux.drivers.dvb/31146/focus=31159

In February Steven came up with initial support for HVR 4000 in the multiproto 
tree.
http://thread.gmane.org/gmane.linux.drivers.dvb/31605/focus=31644

Furthermore at this time the dvb-apps (at least parts of) were started to be 
extended by multiproto support, so that more people (which do not write their 
own applications...) could start testing.
http://thread.gmane.org/gmane.linux.drivers.dvb/31726/focus=31729

In March Steven asked for the status of multiproto. Manu noted that the API 
should be fine, but also asked Steve to look into dvb_frontend where Manu was 
not sure of not having introduced new issues.
http://thread.gmane.org/gmane.linux.drivers.dvb/31938/focus=32144

End of May 2007 still problems in dvb-core, which were related to the new API 
came up and were fixed:
http://thread.gmane.org/gmane.linux.drivers.dvb/33893

Then in Sept 2007 discussions came up how to integrate the multiproto API, 
doing this as experimental or non-experimental.
http://thread.gmane.org/gmane.linux.drivers.dvb/36082/focus=36411

In Oct 2007 Steven abandons his support for multiproto, due to delays caused 
by several reasons. Political, surely also personal, but also technical.
http://thread.gmane.org/gmane.linux.drivers.dvb/36583/focus=36670

At the same time some more sophisticated DVB-S2 featues were requestes by the 
users:
http://thread.gmane.org/gmane.linux.drivers.dvb/36785/focus=36789

Finally in Nov 2007 Oliver did a full review of the new code, which was 
necessary for merging. Still he asked for more reviewers.
http://article.gmane.org/gmane.linux.drivers.dvb/37665

In January 2008 another user-initialised thread came up:
http://thread.gmane.org/gmane.linux.drivers.dvb/38529/focus=38544
Still testing is obviously needed as bugs still come up.

In Apr 2008 VDR announced support for multiproto tree, so that more testing 
can be done by many users.

End of May Manu left for travelling and personal stuff until August, just with 
short breaks apllying some minor patches. Still some users report issues with 
multiproro, which were not fully taken care of.

After his vacation Manu came back on this topic and did another shot at a pull 
request.

---

So this is how I see the history. Still 2 years is a very long time, but 
everyone should keep in mind that introduction of DVB-S2 support has been 
(still is) a big task with many problems. At first of course it is a big API 
extension, which is always problematic.
Furthermore it is an API extension for a hardware which still is not spread 
too widely and especially was not spread in 2006. And even those who had 
proper cards for receiving DVB-S2 still were not able to make any use out of 
the received data. To properly do testing at user side it was really necessary 
to at least have a way to watch some of the distributed content, just to be 
sure it is working well.
This was not possible for a long time due to lacing features in ffmpeg and 
missing alternatives. Still I think the only really working way is using a 
binary Windows codec named CoreAVC.

Keeping all this in mind two years are not too long in my eyes.

So this are just my 2 cents on this topic. All that I am interested in is a 
properly working API with wide application and driver support. Which proposal 
ever fits better - but decided on a technical base and not on historical or 
personal terms.

Regards,
Julian


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/