Re: Code removal - Was Re: Future of the ie(4) driver

On Thursday 05 July 2007 06:45:51 pm Julian H. Stacey wrote:
> John Baldwin wrote:
> > The ie(4) driver in 7.x has several issues.  First of all, it has several 
> > compiler warnings that haven't been successfully fixed in several years 
and 
> > are currently just ignored.  More importantly, it hasn't been updated to 
use 
> > more modern FreeBSD APIs like bus_space (still uses inb/outb) and SMPng 
> > locking.  If someone is using this driver and is willing to test fixes for 
> > it, then it can be updated.  If there isn't anyone who is using this 
driver 
> > and willing to test fixes, then it will be removed from the tree at some 
> > point in the future (say a month or two).
> 
> I reduced "cc: stable_at_freebsd.org, current_at_freebsd.org" to current_at_
> & changed "Subject:" so as not to cross post this tangential reply.
>   ( BTW I checked, I don't have any hardware that uses "ie" )

I've actually had a victim^Wvolunteer in reply to this post.

> What's concerned me increasingly for some time, (& nothing personal
> to any individual, (the above just a useful illustration ) is a
> tendency in FreeBSD for developers to say:
> 	~Unless anyone speaks in [time] I will discard [whatever]~
> Then months later a new release is rolled, & months later users upgrade, &:
> 	"Oh my god! they removed the XYZ I use ! ... Aargh!~
> 
> So when discarding, it seems best to adopt a policy to warn as
> wide a user base as possible, not just developers.  
>    Not just current_at_ or stable_at_ but at least all of hackers_at_.

hackers_at_ is a much smaller audience than stable_at_.  hackers_at_ isn't a user list 
so much as a developer list, whereas users hang out more on stable_at_, so it's 
a far better forum.  I have had success with this approach in the past with 
doing MP safe locking.  I had to resort to threatening removal before someone 
would test patches to lock wb(4) (which someone did do, so the driver is 
still in the tree).  No one ever stepped up to testing the locking patches 
for el(4), so I checked them in and eventually removed the driver.

>    Even then we risk hurting happy users of FreeBSD, eg
>    ISPs etc who just don't have time to read hackers_at_ every day.
> 
>    Maybe FreeBSD should have a low bandwidth mail list, that managers
>    & busy admins could safely subscribe, so they get long warning
>    of functional removal ? Such things as eg 16 bit PCMCIA removal
>    (after 4.11 before 6.*) would have gone to such a list, etc.

More mailing lists that people don't read isn't going to help.  One could 
maybe consider sending out a mail to announce_at_ as a second warning after the 
notice to current_at_/stable_at_.

> Good PR to keep wider user base informed of planned removals,
> & some otherwise unknowing users might then reply 
> 	"OK, I'll install current/ stable on a spare box, & give
> 	developer(s) access, as I can't afford to lose functionality~

This has already worked in the past via the current_at_/ stable_at_ approach (really 
stable_at_ is what reaches the wider audience).

-- 
John Baldwin