[meta-freescale] Pros/Cons of Yocto vs Ltib

Trevor Woerner trevor.woerner at linaro.org
Sat Oct 19 10:48:09 PDT 2013


On 18 October 2013 12:41, Joshua Kurland
<joshua.kurland at adtecdigital.net> wrote:
> Why Yocto instead of Ltib?

Although embedded devices are much more powerful (CPU throughput, MMU
ability, RAM capacity, flash size, graphics acceleration, etc) than
they ever have been in the past, for most of their history (and even
today) installing a desktop distribution onto an embedded device has
either not been practical or is not necessary. So embedded devices
have always needed embedded solutions to this problem. Additionally
most embedded devices require a cross-compiler environment for
development work.

Historically everyone solved these problems using their own,
"home-brew" solutions.

The list of problems associated with the home-brew solutions is quite
lengthy. Although many of them shared various ideas and solutions,
each one still required their own learning curve and had their own
quirks. Even within a given company, different boards/products would
have different (incompatible) home-brew solutions! (due to different
teams solving the same problems in their own ways) As an end-user you
were at the mercy of whatever the manufacturer provided. Some
solutions were stuck at very old versions of various components making
it almost impossible to use newer Linux features and drivers. Others
were so cutting-edge that nothing seemed to ever work. Then, to make
matters worse, some frameworks would simply get abandoned as new
products came to market. Home-brew solutions often required special
host system setup/configuration that would fall outside the purview of
the build framework itself, making it hard to migrate to a new machine
or for a group of developers to collaborate on the same project from
separate development host machines. The list goes on an on... and
often you only ever stumble upon the faults of a given framework after
you have shipped product or are mere days away from release and
suddenly realize some major requirement can't be easily fulfilled.

So the whole embedded Linux development ecosystem was badly in need of
some sort of consolidation.

...and that's where Yocto comes in. Yocto is a project endorsed by The
Linux Foundation, so it tries to be manufacturer/vendor neutral. And
instead of creating or picking an embedded distribution and
standardizing on that, they decided to back a framework that helps you
create your own distribution. Had they simply chosen a distribution
and asked everyone to use it I think fewer members/developers would
have backed that decision than have backed the decision to use a
distribution-creation framework instead. No one distribution is ever
going to satisfy everyone's needs, but if everyone used the same tools
to create their distribution it will be easier for people to
understand and work with it.

So "why use Yocto"? Because it is about time we stop re-inventing the
wheel (from scratch) every time a new embedded board comes out :-)

I feel for your situation. Almost all of the companies for whom I've
worked in the past were replete with managers who preferred to look
behind rather than forward.

There are so many "problems" that Yocto already has answers for (even
if you haven't stumbled upon them yet!) entire manuals could be
written describing them (and more!) :-D



More information about the meta-freescale mailing list