Want to know what I’m telling my customers about QSSL? I tell them, “QNX4
was a fantastic product. It is extremely reliable an lightning fast.
Besides you can get old QNX4 licenses all over the place for a song. Yes,
it is true. QNX4 is sealed in stone and they’ll never write so much as a
new device driver for next years latest and greatest hardware. But if you
can find the hardware that runs QNX4 it’s the bomb. QNX6 is still missing
too many features. But worse than that, it is across the board slower in
every measurable way AND QSSL doesn’t seem to care. Instead they come back
and say ‘yes, we know it is slower but it can run on all these different
hardware platforms’.” Well, guess what. Most developers are only
developing for one platform. I have still only successfully installed QNX6
on systems where I could install in on a Windows partition first. I can’t
even get straight answers on how to install QNX6 onto a system that only has
a QNX4 partition.
To install QNX on any system besides a FAT32 based Windows system you boot the
CDROM and install QNX into it’s own partition. The default for QNX6 is
t79 so it co-exists with a side-by-side QNX4 install. It is possible to
fake out a QNX6-in-a-QNX4 install using the same methods that are used in the
Windows install but that is far from an idea manner to run a “real” install.
Simply put the .ifs of choice (qnxbasedma.ifs) in /.altboot and make sure you
have the qnxbase.qfs and qnxroot.qfs (pinch them over the network from a windows
based install) and put them in /boot/fs on your QNX4 disk. One issue you will
hit is the need to build your own qnxbasedma.ifs that is small enough to
boot with the old QNX4 partition loader, just take out the devb’s you don’t
need. Once you have this working you can easily burn a CD with a script to
do it all, and in the “old days” this was how you installed Neutrino.
As for “Across The Board slower in any measurable way”, would you care to post
your test code, test hardware and results under QNX4 and QNX6? The numbers
I have seen show a small increase in the context swtich time on an idle system
but a far more consistent time under load. The only other place I know enough
numbers wise to comment on is networking speed, and QNX6 is faster then
QNX4. There is no reason for things to be slower and I (along with most of
R&D) would be happy to have explict test cases that are slower (idependant of
hardware) so that we can identify the bottlenecks and remove them.
thanks,
chris
–
Chris McKillop <cdm@qnx.com> “The faster I go, the behinder I get.”
Software Engineer, QSSL – Lewis Carroll –
http://qnx.wox.org/