Fwd: Experiences with KDE / Plasma in an enterprise environment

[Christian Loosli]

Hi David, 

Am Donnerstag, 7. April 2016, 01:25:22 CEST schrieb David Edmundson:

> On Wed, Apr 6, 2016 at 12:40 PM, Christian Loosli <kde at fuchsnet.ch> wrote:


> > Hello list,


> > Thomas Pfeiffer informed me about this list and asked me to post my
> > feedback
> > from an enterprise point of view to it, as we were discussing it on a
> > different list.  Apologies that my first post thus is already filled with
> > (hopefully constructive) criticism.


> 
> Firstly, thanks for the feedback, it really is appreciated and I would
> encourage others to share their specific constructive comments too.

Thank you for the quick reply.



> > Summarized:
> > 
> > 1) Logging in is not possible with a directory service, which is the
> > default
> > in most enterprise environments.


> 
> I owe you a personal apology on that one, I remember you telling me about
> that issue in person, and I promised to fix it.

Meh, try to imagine this report wasn't written by me. If I'd still work for 
the previous employer, I would have used my team account to make it less 
personal. Because: I try to see this from a business, not personal point of 
view. And from a business point of view: I'm afraid that is one of the thing 
that is past apologies, in the business world if this happens it's roughly 
when people send someone to you on site with a heavily overpriced bottle of 
wine and an offer to use their services for free or the likes. 

It's rare that people with some FOSS and technical knowledge are in charge of 
decisions, and if someone with a pure business background sees this, with 
competitors available at the same price, you switch. Because these kind of 
bugs are so terribly bad that the additional workload of switching is gladly 
taken.  And in this case: it is very easy to ditch plasma for something else. 
I've seen co-workers doing it, it roughly takes an hour maybe to configure 
your new desktop to your likes, and you can still use the few KDE apps you 
loved (e.g. still most of us used kate as an editor, despite using Unity or 
Cinnamon or whatnot) 

Also it just displays a certain odd focus. If you have a bug like this open 
for a year whilst features are added, from a business POV you wonder what 
exactly the focus of a certain provider is and how safe you are using that 
provider's products. Especially if it is one which isn't that hard to fix  
(some other themes fix it, fedora ships one by default to avoid exactly this 
bug)



> I did start, but then things happened and I genuinely forgot; I've made a
> bug report now, which will get fixed. I think we got a new design for the
> login screen from the VDG which addresses this so it's going to go away
> soon.
> In the mean time, I recommend anyone to change their SDDM theme to
> something else to resolve the problem

I'm afraid that alternatives are limited. Not taking in everything deprecated 
and things that will pull in tons of dependencies or lack features wrt session 
management, it's only lightdm remaining, which has it's good own set of 
issues.


> > 2) Daily tasks like undocking a notebook lead to the whole session
> > crashing,
> > resulting in a full data loss of all open applications


> 
> 3) Various visual glitches and papercuts all across the applications
> 
> I like to think we're over the worst of that. Most the frequently occuring
> crashers are gone. The rate of new bug reports is going down quickly, and
> the number of introduced bugs is pretty slim.

Sorry that I have to disagree here. A full session crash leading to data loss 
is not the worst being over, and from the papercuts mentioned I can still 
reproduce all of them, visual glitches to bugs, on the most recent plasma, 
frameworks and kde apps releases. 

Also I got reminded of another papercut that annoyed some of my former 
employees: keyboard navigation. Mainly with plasmoids, there are lots of 
things that you can initiate using the keyboard, but then you have to switch 
to the mouse because either the needed controls don't have input focus or, 
worse, the plasmoid simply doesn't allow keyboard navigation. Probably also 
affects some apps, but it's mostly plasmoids that got reported to me.

I am not saying you aren't improving. Plasma, frameworks and apps certainly 
have improved over time. However, the current state I reported is still 
current with the most recent releases and, in my opinion, far from being 
usable in an enterprise environment. 



> I'm not going to pretend we got everything right, clearly we messed up a
> lot, but I'm not sure what we can take away that we're not already doing -
> other than having more manpower.
> I think we should be in a reasnoble state by the time we see the first LTS
> from any distro based on Plasma 5 which AFAIK hasn't happened yet.

Yes, but don't let that trick you. Speaking of experience, LTS is used a lot 
on servers, where there is usually no X11 running at all or, if, then 
certainly not with KDE. For desktops LTS just has terribly old software 
sometimes  (e.g. various libraries, utilities such as git ...) so at some 
point repackaging these gets so annoying that on the desktop you switch away 
from LTS, as we did. And KDE is desktop. 



> > 4) Base concepts that are hard to understand and rarely used
> > 


> I certainly don't want to be going down the route that anything unique is 
> inherently bad.

And I never wrote it is. Sorry if that came across wrong.
It becomes a bit bad when it is forced down user 
throats for daily tasks when it is in general not adapting a lot. Google tried 
the same with one of it's products, the backlash was as expected.
 


> In this case I don't think activities is a problem, the problem is that the
> person couldn't change their power profile without unnecesary steps; there
> maybe needs to be more decoupling there so that common tasks can be used
> independently.
> 
> I'll talk through the power manager example with some people, maybe get
> some ideas on cleaning that up.
> If you have any more specific stories like that, that'd be great to hear -
> as that's the sort of thing we can work constructively with.

Sorry, so far no, I'll try to get in touch with my former employees / co-
workers to see if I can get more. The rest are my personal issues, and for 
these I won't use this list, I already have bugzilla and IRC for that.. I'll 
certainly let you know if I get more or more details, though.


> 
> 5) Unreliable updates to recent versions including data loss
> 
> 
> By data loss, are we still talking purely about the config settings?

Yes-ish. However, for users, their starters, icons and keybindings are not 
perceived as "config settings"


> Has that happened since during the 5.x series upgrades?

Rarely, mostly when one of the missing applications got ported from 4 to 5. 
Side note here: with some plasmoids you have configuration loss without 
updates, favourite example: if you switch between the various task lists in 
the panel (icon tasks, regular) you'll lose the previous settings when 
switching back, e.g. all your starters. I don't think I reported that one yet, 
I shall) 



> I think we definitely could migrated a lot more than we did.
> Turning this into a question (for a wider audience), is 4.x to 5.x config
> migration something that's worth fixing or is it effectively too late?

That's a good question that is hard to answer. I tend towards "for regular 
users it is way too late, but for enterprise it might not be". As you pointed 
out, LTS still uses KDE4  (as do some distributions, e.g. debian, even some 
KDE devs I know still use 4), so there might be some users yet to hit. Also 
not all apps have been ported to 5 yet  (hello kile etc.) or are distributed 
due to dependencies on other 4 libs or apps  (PIM/Kopete, kate related like 
kile ...) so at least for these it would be worth fixing. 

If not: it's definitely a big red reminder for the next transition. Because in 
quite some cases (e.g. keybindings) the functionality to export and import was 
fully there, it just wasn't used. 



> > 6) Response times for pressing issues are way too high, two critical bugs
> > with
> > either data loss or not being able to log in have been reported and known
> > for
> > a year without fix
> > 


> You clearly have two specifc ones in mind. Which ones are you referring
> to?

The initially mentioned not being able to log in (which has it's Birthday on 
launchpad this month) and the full session crash when undocking a laptop, the 
Qt bug leading to it has also been open for months. 



> > [snip]



> > Kind regards,
> > 
> > Christian
> > 
> > 
> > PS:
> > As I am a bit experienced with similar discussions on bug trackers, IRC
> > and
> > other mailing lists, I recommend saving some time and not coming with the


> 


> > following frequently given replies:


> Are we really that bad  ?

Sorry, but: yes. 

See, in most places, including the mailing list which lead to me posting here 
(it's internal, so I won't go into details), you see the same behavior over 
and over again. 

People criticize, then some people start to: 

1) blame others 
2) marginalize the issue  (not specifically talking about your reply here. I 
won't point out specific famous examples, as it will probably sidetrack the 
discussion) 
3) start offending others or being offended, up to the point where they 
threaten to leave the project or delete code (the latter happened on the 
public bug tracker) 

I fully understand this from a personal point of view. It's your product that 
you invested a lot of time and work, paid or unpaid, in. However: this is just 
not professional. You could now say that KDE is not professional, because 
indeed, most (not all) of the contributors are not paid to work on it. 
Question though is: do you want to look professional and have a chance in a 
corporate environment? Because if yes: like that you don't. 

Also if criticism gets that loud  (I remember a famous review and blog post on 
phoronix, I remember longer rants on German Heise, I remember mailing lists) 
it's just not the time to say  "but our tests run fine" or  "but we improved". 
I understand that for developers it is annoying to get so much criticism, but 
trying to create a "everything isn't as bad" atmosphere and asking people to 
tone down (Valid) criticism is not going to resolve things, it might actually 
be bad for resolving things. Also in a professional world: it looks, sorry the 
word, ridiculous. Imagine for once if the big players, say Apple or Microsoft, 
reacting in that way. Yes, they are paid, yes, they do have more manpower. But 
see the first part of my reply: people in charge of decisions simply do not 
care.  Competitors are available for similar costs, they work, ditching is 
easy: you ditch the product and choose a different one, period. 



> 3) It's the fault of the framework we use (Qt):


> > I shall get technical for a short moment here: in
> > case of the screen being NULL which leads to the data loss mentioned in
> > experience 2: Yes, this might be a Qt problem, but you can actually check
> > and
> > avoid it in your end.


> 
> In general yes, I agree with you, we are too quick to blame others, but
> then not really get involved in being proactive reporting it, fixing it, or
> working round it when we can.
> 
>  But given you got technical on that specific bug, I will have to give a
> specific rebuttle. The check you're saying is needed *isn't* on our end.
> See the trace at https://bugs.kde.org/show_bug.cgi?id=360134   frame 7 is
> in Qt's X backend.

I'd have to dig into the code here, but: I'd be _very_ much surprised if it 
couldn't be workarounded.  Worst case: if you rely on a component which leads 
to such a crash, you simply ditch that component and implement that piece of 
logic yourself.  I know that this is a horrible work-around which leads to 
code duplication and maintenance. I know that this is bad for platform 
independence  (even though: we are talking about the screen locker here, which 
takes ksmserver (sp?) down. Said locker is exclusive to X11 / Wayland, so that 
is not as bad as other things where Qt does the platform abstraction). 

However, a crash that bad   (keep in mind: everything you had open during 
undocking, be it libreoffice, kate, kdevelop, eclipse, no matter what: it goes 
down with the session and you lose all your data) merits whatever workaround 
is possible. That is the kind of bug which lets other software providers skip 
their usual patch schedules, here it has been open for ages. (And will stay 
so)


> David

Kind regards, 

Christian


PS: I am not saying that other environments / providers do everything better. 
I've seen  "RESOLVED - Won't fix, user is stupid" on public bug trackers of 
other systems. However, bugs of a certain magnitude get fixed within 
reasonable time there, and I have a hard time believing that some of these 
projects have more manpower than KDE does. 

So: don't get this wrong as  "everything is wrong with $something", but for 
here I'd rather just put what I perceived as making KDE / Plasma unsuitable 
for enterprise environments, not praising the good things