[conspire] could Linux desktop go reasonably any faster?

[Rick Moen]

Quoting Ivan Sergio Borgonovo (mail at webthatworks.it):

> If you're not on mobile (and probably even if you are, unless the
> software is really really badly written) I doubt you'll perceive the
> difference even on not so modern hardware.

You mean Xfce4 and GNOME3?   FWIW, what I saw a few times was hardware
bogging down badly on typical GNOME3 builds (with and without Unity),
that then did not also bog down similarly with Xfce4.  That was my
experience, which I merely offer for whatever it's worth.

Not that I _personally_ care a lot (for upthread reasons), but I keep a
rough sense of hardware needs because I get asked by CABAL newcomers
from time to time.  First question I ask before answering is 'How much
RAM do you have?', which I ask because I'm an incurable optimist and
sometimes a slow learner.  Then, I remember that any answer I get to
that question would probably be erroneous, ;->  and check total system
RAM for myself.  The vaguer the user's knowledge is, in _almost_ all
occasions, the more ludicrously large his/her total RAM is.

I remember one astonishing _exception_ to that rule, back when CABAL was
still meeting at The CoffeeNet in San Francisco (pioneering Linux-based
Internet cafe that I helped build), downstairs from where I lived.  A
fellow brought a dusty old, hulking pee-cee in to have us help 'install
Linux' on it, and I stared at it for a good 20-30 seconds before I
correctly remembered _what_ this hauntingly familiar but mysterious
machine architecture was.  I should stress, this was about the year
1998.

It was an XT clone.

I boggled, I laughed, and then after regaining my composure I explained
to the unfortunate person that, no, CABAL would not be installing any
Linux distribution on his XT.  Yeah, there was indeed this thing called
ELKS (http://elks.sourceforge.net), but... actually, guy, sorry, no,
on reflection, let's be serious, please directly go to Weird Stuff, have
a great time, and get something originating from this decade.

I took care to be really nice to him.  No 'here's a nickel' cracks, nor
any about some boat being short an anchor.

(I have no idea where he got a working XT clone in the late 1990s.
But he had absolutely no idea that it was only good enough for DOS
Wordstar retrocomputing on a good day, and only if he wanted to rely on
perilously decrepit hardware that was likely to lose its magic smoke at
any moment.)


> I'd say main problem is stability.
> But that's because we are in a new moment of transition (touch,
> hybrid mobile UI, wayland...).

You first.  ;->  OTOH:

> So I moved to xfce4 from KDE. It makes easier to use sid.

Yeah, I really rather like Xfce4, too.

> Still no matter what you say about the *nix philosophy, it is always
> a matter of compromises... or we would be running a microkernel
> written in C++ or whatever...

{shrug}  Your system, your choices.

[Akonadi framework]

> I'm speculating... let me guess how it went... we've to manage data,
> calendar, address book seems to be fit for a DB... hey but we have
> semantic desktop, we need full text search and well now we have a
> bunch of applications that may have to access this DB concurrently
> then sqlite is not suited. Let's go for an instance of MySQL for
> every user.

Yeah, they're really big on pervasive search/indexing and have been for
a long time.  I entirely understand the requirements analysis involved;
I'm just wary of the results, because I've seen how that goes wrong -- 
often from the very same desktop coders (well, probably their
predecessor coders).

'Integration' had the same drawbacks of entangled code interfaces and 
near-impossibility of debugging on Win32 desktop computing, too -- 
before the *ix DEs decided to replicate (er, maybe 'port over' would be
an apt metaphor) what looked to me like many if not all of the
Redmondians' errors.  So, I opted out, and classed 'integration' as an
antipattern to be wary of.  

Works for Me.{tm}

> It seems they gave up somehow and nepomuk is dead, replaced by baloo
> (a file search sort of). But I still can't understand why akonadi is
> still around.

Back when NEPOMUK (what did they have against Hummel, BTW?) was all the
rage, The Wave of the Future, Always Has Been, Always Will Be[tm], I
remember people noticed that their systems bogged down under staggering
system loads, and it turned out to be NEPOMUK.  I think that was the KDE
generation where we heard those tales.  Hard to keep them straight.
Hardly is one released when it gets thrown out and rewritten from
scratch.

Recently, there's been some hilarious security attacks against one of
the colossal DEs via the metadata datastore and the search software.
Oh, yeah, that's right, GNOME again:
https://scarybeastsecurity.blogspot.dk/2016/11/0day-poc-risky-design-decisions-in.html
https://scarybeastsecurity.blogspot.dk/2016/12/redux-compromising-linux-using-snes.html
https://scarybeastsecurity.blogspot.com/2016/11/0day-exploit-compromising-linux-desktop.html

The GNOME 'Tracker' (metadata-crawling and indexing software) did dumb
things, and thereby was exploitable.  The maintainer of that software 
ducked and weaved nicely and removed the postings of critics when they
commented on his blog.  (He told them to take it to LWN.net, which is
fair enough.  It's his blog; if he can't be a microscopic dictator there, 
where can he be?)

 

> Once upon a time the PMI suite in KDE was a nice set of applications
> and if Linux had to aspire to world domination it looked something
> that could really take the place of Outlook/Exchange and it was made
> in a way you didn't have to pick the whole package even if picking
> up the whole package gave advantages.

Yeah, never entirely bought that notion.

Chandler, Mitch Kapor's / Open Source Application Foundation's
open-source PIM in Python (GTK+/wxWidgets-based), looked for a brief
time as if it would hit the sweet spot, combining Unix mail with iCAL /
CAP-based (IETF standards) scheduling.  They planned (wrote?) a
CAP-compliant server to work with it.  

It reached pre-alpha, and then something catastrophic happened at the
organisational level behind closed doors, Mitch Kapor walked away, and
the entire effort imploded and dropped off the Internet when the OSAF
funding vanished and the Web site went 'pfft'.

In a corporate Exchange Server environment, I found that Mozilla
Thunderbird (Iceweasel) with the Mozilla Lightning scheduling extension,
and using the DavMail gateway to (in effect) screen-scrape Exchange
Server's OWA interface, worked extremely well -- totally open source on
the workstation side.

And a decade before that, when I was at Cadence Design Systems, I just
punted on the problem and ran guest-OS WinXP Pro under VMware
Workstation 5.5 on my Debian ThinkPad, in order to use Outlook and MSIE 
as necessary in the VM.  This was expedient at the time because Cadence
_very_ intensively used Exchange Server scheduling (I couldn't afford 
to miss meetings because of flaws in reverse-engineering), and also had
a depressingly large number of ActiveX-dependent intranet Web sites I
couldn't live without.  (Even if DavMail had existed in the mid/late
2000s, it wouldn't have been quite good enough, and I'd have still
needed a Web browser able to do ActiveX.)


> I don't think that's a fault of "integration" rather of too much grandeur.

I think we're just describing different parts of the same elephant.
Maybe the delusions of grandeur are an emergent effect of the same
design that aims at integration.  Not sure.  I just note the
characteristic outcomes.


> >>KDE is changing faster than xfce4 and it is made of way more parts
> >>and this has side effects if you're running sid.
> 
> >Yeah, that's one reason I figure, when people claim KDE4 is great these
> >days, I say 'You first.'  ;->  And I've heard way too many horror
> >stories about KDE autocorruption of its files.  Which is, as a different
> 
> But you can't relate KDE failures to the failure of following the
> *nix philosophy.

People often mean different things when they use the latter catch
phrase, leading to long wastes of time while they talk past each other.
That is one reason why I avoid that term in this context.

[VM usage is a killer feature, though.]

> Not in my use case.

You would know.  

I previously considered it a niche of sorts, a superior alternative to
dual-booting for those who _actually_ need multiple OSes.  But lately 
I've started to see many additional uses.

One is monitoring/managing security of a host:  It's much more easier
for a host OS to manage/monitor a guest OS than the reverse.  And also,
it's very difficult for a rogue guest to attack the host.  

Another is prototyping & migration.  Run your production host in one VM,
and the beta of the replacement in another.  On flag day, swap the beta
into production, start builing the next beta VM.

Or say you want to design and test an entire Hadoop cluster.  You
_could_ have six or so rackmount machines in your garage wailing like
banshees and chewing up power.  _Or_ you could pay a ton of money to
Amazon and do it on AWS.  _Or_, you could do it in six VMs all running
on your laptop.

Start thinking creatively about VMs, and you might be startled to find
that there are indeed uses in your use case. 

Or maybe not.  You would know.


> I rarely see people using their phone on a desk.

But many people seldom use their desks, as they are often away from it
and using their smartphones or tablets instead.

Workstations and laptops aren't dying; but they're no longer the centre
of the world, either.

> Waiting Ryzen. I hope at least for some more competition.

You know, you're the second extremely well informed person to say
something like that.  (/me also waves to Dana.)