Re: a joint letter on low latency and Linux

[Paul Barton-Davis]

>Choose a few distributions to "support" and
>supply packaged kernels for them. Updating
>the kernel for most distributions is simple
>enough for average users. Now you aren't
>asking end user to patch+compile the kernel,
>just install an update. Some of the distributions
>may even include the patch in their kernel releases.
>(RedHat usually has a large patch list to the stock
>kernel to fix bugs or add drivers.)

This might satisfy many of the current generation of potential users
of low-latency audio/video/MIDI/game apps, but it will do very little
to help the set of people that I'd like to appeal to.

More importantly, it does nothing to help convince existing commercial
software developers that Linux really is a better platform for this
stuff (which it clearly is, once the patches are in place).

  Me: "Linux has much better latencies than Windows, and it also has ..."

  Steinberg (for example): "Which version of Linux ? We've heard it
                            has 100ms worst-case latencies"

  Me: "Well, sure, the stock distributions are pretty bad, but users
       who want this stuff can get a fix for that that improves
       things, and you could even include it with your apps."

    <click>

>The apps you release coud handle all of this
>in the installer. Just write a smart installer
>that doesn't do bad things to the current system.
>It could install the new kernel, 

which means checking that they are running a suitable set of
kernel-related utilities, and possibly updating them, etc. it may
alter other behaviour - security issues, "fake" filesystems that are
not in their /etc/fstab's, etc, etc. the list of issues goes on for
quite a while.

its not suprising to me that very few (if any) groups/companies are
trying this at this point in time, but instead just go down the route
of preparing entire new installations.

--p