Igor Kovalenko <Igor.Kovalenko@motorola.com> wrote:
Fairy tales again. Didn’t you guys gave it a thought before you
announced upcoming lxrun availability and started all the buzz about
embracing Linux? Or was it other way around (announce first, then think
if we really can do it)?
It was started as a home/side project as a proof of concept.
Funny thing is, in either case you look bad because you are either
unable to estimate realistically complexity of a task or do not bother
to do it at all. References to ‘a developer who worked on xxx’ (and then
left I presume) create a feeling that your R&D management does not
really plan anything. A guy comes, he thinks he can do something and you
announce it, he leaves and you’re unable to keep your word. Sounds very
satisfactory indeed.
No the developer is very much still with us. The project was put
together (if I remember correctly) in about a week of time initially
(remember that Christmas holiday last year …). It was used as
I mentioned as a proof of concept, but also to check and see that
the claims we were making about linux compatability weren’t that
far off. By building the emulator we were able to better gauge
the differences between Neutrino and Linux systems and where the
difference was minimal bring Neutrino more in line in order to be
“more like traditional Unix” systems. This was always a problem
with QNX4 in that it had a lot of peculiarities about it.
At the height of the development (just under a year ago) it was
capable of running X with WordPerfect and I think even Star
Office too. The downside is that it required a red-hat install
for most of the libraries (ie a huge overhead).
Like I said this was a home project and a proof of concept, and
certainly not something that others wouldn’t be able to pick up
and run with if they started with the sources again today, even
if they started from scratch.
Speaking about lxrun, NTO isn’t the only kernel being constantly
evolved. SCO and Solaris evolve too (and they support lxrun). In fact
the NTO kernel is probably being evolved less than other Unix kernels
since there’s not much there to evolve. Most of Linux kernel calls would
be libc calls in Neutrino and most of those libc calls would be just
wrappers for message passing to resource managers. While your resource
managers evolve indeed, the interface should be more or less stable, am
I wrong? So why don’t you stop looking for reasons not to do what have
been promiced and leverage those architectural advantages you are so
fond of?
It isn’t so much about kernel evolution, it is about spending time
developing better performance, optimization and loudly requested
features (Unix domain pipes?) so that an application like lxr will
run even better.
It wasn’t my project incidentally so I’m not in a position to say
when/if the code will be made publicly available. If someone
wanted to start on this on their own and asked for help however,
I don’t think that anyone here would refuse to give assistance.
Thomas
.
