[lkml]   [2008]   [Sep]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [linux-dvb] Multiproto API/Driver Update

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

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:

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:

Immediately followed by Johannes stating that he is not satisfied with the API
yet and suggested a rework:

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.

In February Steven came up with initial support for HVR 4000 in the multiproto

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.

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.

End of May 2007 still problems in dvb-core, which were related to the new API
came up and were fixed:

Then in Sept 2007 discussions came up how to integrate the multiproto API,
doing this as experimental or non-experimental.

In Oct 2007 Steven abandons his support for multiproto, due to delays caused
by several reasons. Political, surely also personal, but also technical.

At the same time some more sophisticated DVB-S2 featues were requestes by the

Finally in Nov 2007 Oliver did a full review of the new code, which was
necessary for merging. Still he asked for more reviewers.

In January 2008 another user-initialised thread came up:
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


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.


 \ /
  Last update: 2008-09-19 13:01    [W:0.101 / U:45.504 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site