Re: Re: RFS: socketcan-utils
> On 31.05.2011 09:46, Markus Becker wrote:
> >> On 30.05.2011 09:16, Markus Becker wrote:
> >>> Well, we explicitly need isotp, which is not included in stock debian
> >>> images. That's actually the reason for creating the images. Any idea of
> >>> how to get Linux mainline isotp kernel module into the Debian kernel?
> >>
> >> My problem is, that we're using my isotp stuff for many purposes for
> >> more than two years now. I'm not far from sending a patch on netdev-ML
> >> - but i'm still missing some feedback from isotp-users (like you), if
> >> they think the protocol and it's API are mature enough for mainline ...
> >> is it?
> >
> > Within the next two month, we will test two telematic devices against
> > socketCAN ISO-TP and raw or J1939 frames. The Debian packages were
> > created to get our test equipment ready. Will report to you on our
> > experience.
>
> Thanks! You are also working with j1939 ?? Did you check out the work of
> Kurt van Dijck in socketcan/branches/j1939?
Not yet, but wanted to test it.
> > If sent now upstream, this would possibly end up in 3.0.1 or whatever
> > number the first kernel after 3.0.0 would have, or would it be 3.0.2?
> > Then I could get rid of the socketcan-driver package again, or the J1939
> > would be included then...
>
> If so, it would get into 3.1
>
> >>> Of course one needs to rmmod all CAN kernel modules and load the
> >
> > appropriate
> >
> >>> svn ones.
> >>>
> >>> I would prefer to put an explicit warning into the package description.
> >>
> >> Better we get isotp into mainline :-)
> >
> > Until then, I think I should clarify this in the package description.
> > E.g. some text like this:
> >
> > WARNING: This package contains experimental kernel modules that might
> > conflict with the mainline kernel modules. You are required to unload
> > all provided CAN modules, before using the kernel modules that are
> > created from this package. If you do not intend to use ISO-TP or J1939
> > kernel modules, you do not need this package.
>
> Ugh, that's hard stuff. I don't think it's a good idea to provide 'pre
> release' protocol implementations & APIs, people build their application
> on.
>
> It's not only the mix of modules but the release of an API in discussion.
> Of course you can build a debian socketcan-modules for your own purposes,
> but having this on a 'real' debian repo doesn't look like a good idea.
Well, OK, then I will keep socketcan-drivers private, so that we can easily
install them (until they are in mainline kernel) and only push socketcan-utils
for inclusion. If/when a sponsor is interested.
Markus
> Regards,
> Oliver
------------------------------------------------
| Dipl.-Ing. Markus Becker
| Communication Networks
| TZI - Center for Computing Technologies
| University Bremen
| Germany
------------------------------------------------
| web: http://www.comnets.uni-bremen.de/~mab/
| mailto: mab@comnets.uni-bremen.de
| telephone: +49 421 218 62379
| building: NW1 room: N2260
------------------------------------------------
Reply to: