[LEDE-DEV] Kodi
Daniel Engberg
daniel.engberg.lists at pyret.net
Wed Mar 15 09:31:09 PDT 2017
> On 03/15/2017 11:57 AM, Daniel Engberg wrote:
> > Hi,
>
> Huhu,
>
> >
> > While I applaud your achievement I'd don't see such projects viable in
> > terms of maintainability and longevity.
>
> Applauds from me as well! I started doing so with XBMC a couple of
> years
> ago and eventually gave up..
>
> >
> > Pulling in Kodi will result in lots of external packages and
> > dependencies to make it usable in a reasonable way.
>
> Sorry, but I totally disagree.
> I always saw (and still see) OpenWrt/LEDE as a toolchain and Linux
> embedded distribution - *not* just for routers.
>
> Me and few others already started very early to get it running on
> rather
> unusual devices (landline phones, mobile phones, digital picture
> frames,
> even (mini) notebooks (OLPC, Qi NanoNote, etc.)).
Sure, you can do "novelty" stuff but unless there's a commitment by
several
people you'll end up with a rather hackish (in most cases) approach and
code rot in the end due to lack of interest of original committer(s).
Not
to forget someone(tm) needs (I guess you could call it good practice) to
take care about the code because it's in the tree even if current devs
doesn't have the hardware and/or interest. As far as I can tell nothing
of
what you mentioned is in the main repo and for the very reasons above
most
likely or other developers plainly refused to include it in the tree
from
the beginning.
I'm not saying it should never expand, however going into a very niche
"market" where you already have very good alternatives and code
infrastructure isn't something you want to do with 1 or 2 devs even full
time. I do consider LibreELEC very competent at what it does and
regarding
my concern feel free to have a look at the package repo and compare it
to
what we have currently. Of course, not every package is needed but I
hope
you get my point.
https://github.com/LibreELEC/LibreELEC.tv/tree/master/packages
>
> Almost a decade ago by now we ported Xorg and all it's dependencies to
> OpenWrt/LEDE. Later on GTK, enlightenment and its EFL stack, Qt and
> what
> else. Although the xorg-feed is not in a good shape anymore (and I
> rather discourage anybody from using Xorg if there's no real need for a
> windowing system), lot's of stuff remained, is still maintained and
> used.
> The new video feed is supposed to be kind of a reincarnation of the
> outdated xorg feed.
This is pretty much my point, it never got much interest because lack of
users (possibly several reason for it etc) and the ball stopped rolling.
>
> I highly encourage approaches like those and don't see any need for
> forcing OpenWrt/LEDE being only a router distribution.
> This only counts of course, if one approach doesn't interfere with the
> other. So let's come to your concrete issues:
>
> >
> > There are several issues with this:
> >
> > * In general binary size > performance and/or functionality pretty much
> > always takes priority, this is a major issue when it comes to multimedia
> > (ffmpeg and friends comes mind).
>
> I fail to see in which way additional packages - in this case:
> dependencies for Kodi - do add size/performance issues the existing
> device configurations.
While we still lack a great deal of packages for Kodi there's a quite
high
resistance in doing performance enhancements while increasing the size
of
packages (ffmpeg comes to mind). Adding to that there's little to no
desire
doing platform specific configurations.
>
> > * It's more or less duplicating work already done by projects like
> > LibreELEC, OSMC etc.
>
> Then all work for OpenWrt / LEDE is duplicated work already done by
> Gentoo (emerge build instructions), Debian (deb rules-file),
> OpenEmbedded, you name it...
I honestly think you see my point here, in this particular case.
Please refrain from dumbing down the discussion.
>
> > * You'll need to import a lot more of 3rd party libs and applications to
> > make it a viable/reasonable user experience and there's a overall
> > concern about build performance of the buildbots.
>
> I heard that buildbot performance argument quite often. Back then, I
> eventually gave up, excluded the xorg-feed from the default feeds.conf
> and problem solved.
> Funny enough however, the discussion comes up every now and then again.
> Last time - without me re-initiating it nor taking sides - quite a few
> people who're participating at OpenWrt/LEDE for a long time by now,
> tried to encourage me to migrate the video feed into the packages feed.
>
> Right now I'm fine with having the extra video feed and maybe will open
> up the discussion about having it enabled by default some day. So, no
> additional build time (for now).
This is an issue, recently nbd and jow (if I missed someone sorry) put a
lot
of work into getting package builds more efficient. It's better but as
far
as I know there isn't much headroom. You can read about the
infrastructure
here:
http://lists.infradead.org/pipermail/lede-dev/2016-December/004786.html
I'm sure jow, nbd and others can fill you in about the current state.
>
> > * Will anyone properly maintain all packages? We're already now
> > struggling keeping the current repo somewhat up to date and
> > maintainable, adding very niche packages certainly won't help.
>
> This is indeed arguable, but come on, the situation got *a lot* better
> over the past years and we can always kick out packages which are not
> maintained anymore. No need to deny porting efforts because some day it
> might not be maintained anymore...
>
This is already a problem in the current and the external OpenWRT
package tree to some extent. Unless we have a bunch of people ready and
up
for the task it seems like a very bad and overly optimistic idea.
> >
> > Whether we like or not about 90% or so of all efforts goes into
> > something network related, I don't think we in a foreseeable future want
> > LEDE to turn into a swiss army knife and/or jack of all trades.
>
> I agree on being skeptical about features which would interfere with
> what OpenWrt/LEDE is currently mostly used for.
> But discussed project ain't one of them.
>
> So please let's not deny those effort, discourage projects like those
> and the people behind - but rather support and encourage them in
> up-valuing OpenWrt/LEDE as a whole.
>
> I honestly don't understand the fear here, but maybe I'm missing or
> misunderstanding some points(?).
>
> If concerns however outweigh the benefit I see, I'm more than happy to
> host Kodi and its dependencies in the video feed and risk it not being
> enabled in feeds.conf.default anytime soon.
I'm all for offering it as an external feed as long as you don't clutter
the main repo(s).
>
> >
> > Best regards,
>
> Likewise
>
> > Daniel
>
> mirko
>
Best regards,
Daniel
More information about the Lede-dev
mailing list