Re: the push to get rid of CONFIG_VT in the kernel and the future of Speakup

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Date: Thu, 09 Oct 2014 09:23:51 -0500
From: "John G. Heim" <jheim@xxxxxxxxxxxxx>

Maybe we should change the discussion toward trouble shooting techniques.

Suppose you are the one and only linux systems admin where you work. You 
come in one morning and there are 8 people at your office door saying 
they can't get to their email. You ping the email server and get 
nothing. What do you do next?


	Hello.  I've been a Unix sysadmin for over 20 years and I've never
worked with servers where I require Speakup in the kernel.  In addition,
I've been confronted with John's dilemma many times.  Here are some of the
ways I've tackled the problem, with various kinds of Unix and a wide
variety of machines:

1.  If the server has a serial port, even if it's not used as the console,
I keep a special version of bootable media around that will let me boot the
OS in an accessible form, using the serial port as console, thus giving me
the ability to look at the crashed system and find out what happened and
why it won't just reboot and run again.

2.  If the serial console is not an option, I can usually get the system to
come up  by simply rebooting it.  (most often, if it's not completely
crashed, logging in on the console without speech and shutting down
manually works while, if it is crashed, a power cycle will often do the
trick.)  Failing that, I usually can employ  a modification to option1
whereby I boot from alternate media which I know will get the network up
and running enough for me to get in and diagnos the issue that started this
chain of events.

3.  If those two options don't work, grabbing one of the 8 people banging
on the door saying their mail isn't working for 5 minutes and having them
serve as a 5 minute reader will usually tell me what's  wrong and how to go
about fixing it.  Sometimes it's a quick command that getts things going
and sometimes its more involved, but most of the time, if it is more
involved, I can do something quick to get me going independently, with the
reader's help, and take it from there.  When faced with this situation
where a large number of people are affected, I've rarely heard people
complain that I had to borrow their eyes for 5 minutes or that they wished
they could find someone who could fix things without their help.

	The most important thing I do, however, is try to plan for the
eventuality that John describes.  One of the ways I do that is to perform
system reboots when ever changes are made to the server that could affect
its ability to boot.  These reboots are performed under controlled
conditions and when I know I can get to the machine and get to it via one
of the methods described above.  This technique, combined with a very
conservative upgrade process, gives me a high degree of confidence when I
reboot a server that it will make it up enough to let me in via the network
and thus be able to fix any issues that might crop  up.

	I am a big fan of IPMI enabled servers and so, when possible, when
replacing a server, I'll advocate getting one that has IPMI capabilities.
this gives me access to almost all aspects of the server and, in fact, lets
me build it from many miles away in  most cases.  However, John's point
about not having IPMI available everywhere is well taken and most of my
experience is with environments where IPMI was not available.  Now that it
is more readily available, I embrace it fully and without reservation.  IPMI
has a learning curve, but it's well worth climbing and I recommend it
highly.

	There are those who will disagree with my choices and will have their
own way of doing things; that's ok.  What I wil  say, however, is that a
screen reader that works entirely in user space is a nice thing to have,
and, given the number of people we have working on and using Speakup versus
the number of people orking on and using more main stream portions of the
kernel, including the console driver, I think it's reasonable for folks to
be nervous about including Speakup in their production kernels.  Having a
user-space screen reader that's just as capable, or even more capable in
many cases than Speakup, would go a long way toward making folks feel
comfortable including it in their distros.  

	Many many machines today, including small embedded machines, don't hav
consoles available, even for sighted folks.  In most cases, the sighted
folks use tricks similar to mine to get access to crashed machines witout
consoles.  My point here is that the console, as we know it, is going away
and we, along with everyone else, need to learn how to deal with that new
reality.  One way is a user-space screen reader.  I use Yasr all day every
day, and hav done so for 7 years.  I like it and it works well for me.
There are folks here who don't like Yasr; that's ok too.  Having another
user-space screen reader available won't offend me and, in fact, I think it
would be a good thing to pursue.

-thanks
-Brian

_______________________________________________
Speakup mailing list
Speakup@xxxxxxxxxxxxxxxxx
http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup





[Index of Archives]     [Linux for the Blind]     [Fedora Discussioin]     [Linux Kernel]     [Yosemite News]     [Big List of Linux Books]
  Powered by Linux