On Tue, Jan 16, 2001 at 02:24:51PM -0800, Adam J. Richter wrote:
> Which specific programs would the change to
> /proc/sys/dev/parport/parport0/autoprobe break? I am not aware of
> any that currently use this facility.
Red Hat Linux's kudzu does, for one, although it doesn't really act on
it very much yet.
Presumably Lothar also does a similar kind of thing.
Basically any kind of automatic printer detection/setup utility.
It's not like this can't be done in user space anyway, in case the
length field is absolutely needed.
> If anything, now is a better time to make changes than
> before, since the "code freeze" is now over. I have delaying
> pushing to get these changes into Linus's kernels since the
> "prerelease" series. I do not think that developer who try
> to accomodate this process should be penalized. Otherwise, they
> will probably adjust their behavior accordingly in the future.
All the other changes are good, and don't affect compatibility
(noticeably). It's just the length field thing. I just think it's
too late. MHO.
> What makes the two length bytes "incompatible"? Which programs
> does it break? The USB ioctl returns the length bytes, so my version
> keeps the format consistent, making it a bit easier to write tools
> that use the IEEE 1284.3 facility. Can you explain why we need a third
> format?
Just so that existing utilities know what 'autoprobe' is actually
supposed to contain.
At the moment, it contains something that can be parsed as an IEEE
1284 Device ID, without a length field (with the length implicit in
the number of bytes you can read from the file).
With your change to make it re-read the device ID on demand, it still
contains something that can be parsed as an IEEE 1284 Device ID, so
there's no real format change. As soon as you put the length field in
though the format changes.
Arguably, all of this should be done in user space anyway, but I'll
leave that debate until 2.5.x.. ;-)
Tim.
*/
-- To unsubscribe, send mail to: linux-parport-request@torque.net --
-- with the single word "unsubscribe" in the body of the message. --
This archive was generated by hypermail 2b29 : Tue Jan 16 2001 - 17:54:04 EST