Mailinglist Archive: opensuse-factory (602 mails)

< Previous Next >
Re: [opensuse-factory] Re: [PLEASE SPEAK UP] Disabling legacy file systems by default?
On 2/1/19 8:19 PM, Jim Henderson wrote:

Well, as it happens, I ran into this recently. I got a new microSD card
that was formatted exfat, and I got whatever the standard message was.

My solution - STFW for what was needed to run exfat. Search the repos
for anything exfat related, and install them. Then try again and look at
the logs.

And guess what - I got it working. Took all of about 15 minutes.

That's what I did when I first encountered ExFAT on an Ubuntu box, too, yes.

But ExFAT isn't blacklisted, is it? That's different. IIUIC then even if
the driver is there, it won't be loaded.

I think people should be able to figure it out, yes. Why do we assume
that people are stupid or unable to ask questions when they run into
problems?

Because, for the most part, people _are_ stupid. Even if they're very
smart and educated in one area, that area probably isn't Linux, and if
it is Linux, it probably isn't *SUSE.

Isn't that what the community is for?

No, I don't think it is.

I mean, as an example, I got a new PC recently at work. It has nVidia
graphics. Getting the nVidia driver working on Tumbleweed was pretty
horrible. After a lot of experimentation, I ended up downloading the
file direct from nVidia US and installing it, because unlike anything in
the repos, AFAICT, this does stuff like blacklisting the "nouveau"
driver for you.

Then I discovered that every time TW gets a new kernel, X.11 falls over.
I installed DKMS but it's not working.

So I wanted to try the built-in drivers.

Problem #1: there are 2. One ending in G05 and one ending in G04. I
can't find anything official or canonical with Google as to which
package installs which GPUs. At least nVidia's website tells you that
for its drivers -- but not for *SUSE's.

This took many retries, reboots etc,. as you might imagine.

It's G04. It's a new machine from early this winter, with a new graphics
card, but it's the "older" (?) driver.

I found some changelogs from a couple of years ago that mentioned this.
That was all and it took a lot of digging.

No wonder I couldn't get the G05 drivers working.

And of *course* they don't fail with any helpful error like "this driver
does not support this GPU".

No, it's a random message about no screens found or something.

Problem #2: there are an interrelated complex of drivers to install.
It's not one package and they don't seem to depend on each other.

I needed to manually install 4 separate packages:

nvidia-computeG04
nvidia-gfxG04-kmp-default
nvidia-glG04
x11-video-nvidiaG04

The relevance of this?

Drivers are _hard_, stuff like blacklisting stuff is harder, and
troubleshooting it is not easy.

On a more mainstream desktop distro, I'd run the proprietary-drivers
tool, it would detect what I need, offer me a choice of versions and
just do it for me.

But *SUSE is not like that.

I think that:

[1] We should try to make it more like that.
[2] We should not blacklist code that people active in the community are
using.

It's not common sense that if you have a problem that you should look at
the log files?

No!

You know why?

Because that's not how it works on Windows and that's all most PC users
know.

Second reason: because we have blasted systemd now and as a result of
that, half the most common basic logfiles have bally well disappeared
and are now in some binary log format I don't know, hidden I don't know
where and only viewable by a tool I don't know.

So when I started troubleshooting TW, a very early step was to install
some package that disables systemd's binary logging and gives me actual
logfiles again. *Then* I had to wait for various problems to occur again
so that now I could actually look into plain text logfiles with ``less''
and find out what was going on.

This took extensive googling.

I do not claim to be a *nix genius, but I do submit that this kind of
thing is _not_ "common sense".


Then we need to educate users how to perform really basic
troubleshooting.

"This bridge... how many lanes do you want?"

http://www.jokeindex.com/joke.asp?Joke=3529


Because if they're having problems with this, IMNSHO, then they're having
problems with things like how to log in and enter their passwords
correctly.

You think?

https://en.wikipedia.org/wiki/List_of_the_most_common_passwords

Did you know, the old ICQ chat system still works? It's about the last
man standing. AIM is gone, which even Apple iMessage once used. Yahoo
has gone. MSN IM has gone. Facebook and Google have abandoned XMPP.

But ICQ got sold off and it still works.

But people aren't going back to it. I've asked my friends why not. Some
are on my "buddy list" and it's one of the few person-to-person chat
systems that still works in proper client apps such as Pidgin.

Why not? Because they don't know their old passwords, they don't have
access to the old email addresses they registered under, and so they
can't get their accounts back.

These people are often technies, professional sysadmin types.

Every online service in the world is _littered_ with multiple dead
accounts for people who changed email and forgot.

I still have the same email address as I did in 1991 and I have a system
for generating unique passwords in my head, so I can work out what
password I used, even 25 years later.

But I am very unusual.

So, *yes* not knowing how to log in as another user, how to remember 2
different passwords, stuff like that is hard! Yes, it really is, and
most ordinary users can't do it.

And as I have also said, if we want to get people to come and use our
server distros, then we _need_ the mindshare that comes from having a
good, easy, solid, polished desktop distro to tempt them in.


Using dmesg and other logging systems should be second
nature.

[1] Maybe it should, but it isn't.
[2] It isn't going to be, either.
[3] If of course it actually still works. Systemd or the like might have
replaced such legacy tools, same as ``ifconfig'' doesn't work any more,
or /etc/resolv.conf doesn't.


It's like looking at the event viewer in Windows when a service
doesn't start.

True story. 4y ago I interviewed with IBM for a midrange tech support
job, 2nd/3rd level.

The first interview question was: on a Windows machine, where would you
look for error logs?

This is a question they use to weed out _advanced_ techies from normal
techies.

Yes, really. Not a friend, *me*.

Other examples: you want to test a connection to another machine by IP
address. What commands might you use?

Again: 2nd/3rd line support. I was some 46 years old at the time.

This is _not_ basic stuff any more.

Yeah, my mother wouldn't know to do that - but she would know to ask
someone who knows more, and that's what the community is for.

We should spare the need to ask anyone anything as much as possible.

But we're talking here about a change that affects more technical users -
my mom isn't going to run a combined OS/2 and Linux box and need hpfs
support. People who run that kind of combination are not causal users,
they're people with a specific need and a specific skill set. We
shouldn't treat them like idiots.

Doesn't matter.

Another example:

While at Red Hat, I complained that the Fedora installer failed with a
multiboot setup of Windows and another Linux distro.

Response: Will Not Fix.

Reason: We are a server distro. Servers do not dual-boot. We do not
support dual-booting except with a single copy of Windows installed with
default settings. You attempted to install on an unsupported configuration.

Big companies often feel they don't need to support little edge cases.
But little edge cases that are obscure to 1 person/group/community are
core to others. But to others, these may be prime, high-priority needs.

You shouldn't dismiss anything as unimportant as it's an edge case. One
person's edge case is another's basic spec.

And those users, as I stated above, have some technical chops because
they have a specific need for that combination. I mean really, we're not
just talking OS/2 fans. We're talking OS/2 fans *who dual boot* and need
to support hpfs in order to access data shared with a local OS/2
installation.

That's a pretty small audience no matter how you slice it.

That's not the point.

Stuff like this highlights problems that may be niche when the offending
driver is $FOO but then 6 months later you find it also affects driver
$BAR and some company is shipping a million devices a day that need
driver $BAR. You have 6 users who need driver $FOO but 60 million users
who need driver $BAR. If you'd fixed the problem for $FOO then you'd
have had 60 million happy campers, but you didn't, because you thought
it only affected 6 people and as a direct result you have a major PR
disaster on your hands.

Example: the Pentium FDIV bug.

The specific numbers were super-rare and many people would never have
been affected, but an oversight nearly bankrupted Intel.

Heartbleed.

This stuff really happens.

Don't leave edge cases because they're obscure or niche because it's a
law of nature they'll come back and bite you.



--
Liam Proven - Technical Writer, SUSE Linux s.r.o.
Corso II, Křižíkova 148/34, 186-00 Praha 8 - Karlín, Czechia
Email: lproven@xxxxxxxx - Office telephone: +420 284 241 084


--
To unsubscribe, e-mail: opensuse-factory+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse-factory+owner@xxxxxxxxxxxx

< Previous Next >
List Navigation
This Thread