Kristian Rink posted on Thu, 07 Feb 2013 20:31:25 +0100 as excerpted:
> Hmmmmmmmmmmmmm... So hope folks will be forgiving with me asking this
> question: Is actually anyone _really_ consciously using nepomuk and the
> semantic desktop? No offense, as I like this technology yet so far
> failed to really make meaningful use of it. Browsing the web however, it
> seems weblogs and web sites are nothing short of bashing nepomuk and
> outlining how to disable the "semantic stuff" in order to gain
> performance. The same time, KDE applications seem to just partly make
> use of what nepomuk can do. Is there any reason for that?
Well, as I said with the comment that triggered this subthread, I have
USE=-semantic-desktop set here on gentoo.
Of course one of the biggest gentoo features is the fact that since it's
scripted-build-from-source, they can and do expose a lot of the pre-
compile-configure-time-feature choices to the user (with USE flags the
method of doing so), allowing the user to choose which optional features
to build in, where binary distros have to make that choice at distro
build time, and most general-purpose distros thus end up building in all
sorts of features that maybe 10% of their users will actually ever use.
(With some things, support for obscure media codecs, etc, it's probably
well under that.)
So when I say "disabled", it's gentoo-level disabled, as in, choice made
pre-build, support not compiled in at all.
And while I disabled semantic-desktop and etc with kde 4.7 and its
performance is said to have improved some since then, and even tho I had
it mostly run-time disabled before that, I was **VERY** shocked at how
much better kde performance was, once I actually turned semantic-desktop
support off, rebuilt kde without it, and uninstalled all the now
unnecessary components (including nepomuk, soprano, akonadi, the mysql
and virtuoso database backends, rasqual, redland... strigi installation
is still required for the headers, but I don't have any backends for it
to use so it's hard-disabled).
I mean, I considered kde4 pre-release quality until 4.5. With 4.5, I
thought it finally reached proper release quality and compared reasonably
to 3.5.9/3.5.10, but it was only when I built kde4 WITHOUT semantic-
desktop support enabled at BUILD TIME, that I really saw it speed up back
to what I remembered from the late kde 3.5 era, and I was FINALLY more
satisfied with kde4, than I had been with kde3.
Seriously. I felt like the MSWormOS user that just got the malware
cleaned off their computer and realized how much it had been dragging
things down. Or like I just bought a CPU upgrade of a couple more cores
or a minimum 500 MHz. That's the difference it made!
Which is definitely ironic, since the whole semantic-desktop thing was a
major bullet-point feature of kde4.
But I really AM convinced, that a good portion of the complaints about
kde4 speed, compared to kde3, at least once the big bugs got worked out
by 4.5, is because of the support for semantic-desktop that's dragging
things down even when run-time disabled to the extent possible.
And IMO it's really a shame that pretty much the only folks who will ever
get to see kde4 run as I now know it can, are the folks that build from
source and disable semantic-desktop at build-time. I really do think
there should be at least ONE binary distro that builds kde4 with that
stuff turned off, so people can actually see how fast kde4 can be,
WITHOUT having to build it themselves. It might not be that many that
would choose to go that way, but it seems to me that at the moment
there's an unfilled niche, just inviting some distro to fill it.
Meanwhile, I think the real market for the whole semantic-desktop thing
are the sort of people that have a desktop full of icons, because that's
the only place they ever save anything, and if by some chance they save
it elsewhere, they never can find it. The whole tagging thing. The
whole timeline thing. It's for people that don't "think like a computer"
and aren't interested in LEARNING to "think like a computer". If they
spend another 3-5 minutes doing a task because of the drag of the
semantic desktop stuff, no big deal, because if it weren't for that,
they'd have spent 10-15 minutes fighting the computer, trying to get it
to do what they want, but not knowing how to communicate what they wanted
to do, because they simply don't think the way a computer does.
By contrast, the folks already reasonably comfortable with a normal
computer don't need the stuff the semantic-desktop brings. Sure, they
find it /nice/, even something they might find /useful/, *IF* it didn't
have that terrible performance cost. But they're already used to
thinking in terms of file hierarchy and if they don't know EXACTLY where
they put something, they know it's either in this directory or that one,
and can find it within a couple minutes at most. Worst case, they can
run a real-time file-search (aka grep) for it, pointed at a specific
point in the filesystem tree that's small enough that it only takes a
couple of minutes to get a result -- searching in real time.
They never have the problem of not being able to find that file they just
saved, because they knew what the file was, and naturally organize their
filesystems in ways that make sense to them, so they can find their
content pretty fast.
As such, the added value of all that fancy semantic-desktop stuff is much
lower for them, while at the same time they tend to appreciate the
tradeoff in performance it costs, and are thus rather less willing to pay
what to them is a higher cost, for what to them is a far lower benefit.
But get that guy or gal that can never find the file they just saved,
turn them on to the stuff-it-all-in-one-dir-and-use-tagging-and-
timelineing of the semantic-desktop, and once they get the hang of it,
they're not going to want to give it up, regardless of the performance
cost. Because for them it's a massive improvement from the way things
were before, as they're no longer having to fight the computer every inch
of they way, to get done what they want to do.
Now consider, how many of those computer-technophobes are going to be
blogging or writing forum posts about how well this new semantic-desktop
thing is working for them? Now compare that to the computer-technophiles,
like me and I guess like you, who are already comfortable with computer
technology as it was, and thus find the performance-cost-price of the
semantic-desktop stuff far higher, for a far lower benefit they get from
it.
Self-evidently, it's a technical medium, and the technophiles are going
to be the ones using it to tell the world what they think about stuff,
including the semantic-desktop. So the sample bias is pretty huge, and
it should therefore come as little surprise that there's few in the
technical sphere out there singing the praises of this new technology,
which to them is simply too costly for too little benefit, while the
people that find it REALLY useful... simply aren't going to be the ones
in the forums and blogs and etc talking about it.
Meanwhile, from a historical perspective, it's also worth noting that a
lot of the initial semantic-desktop work was done under grant-sponsorship
from a couple European governments. That grant money is now of course
long gone, and with it, all the people investing all those coding hours
in the implementation.
But it wasn't just kde that got grant money for those projects. Several
other projects got grant money as well. And significantly, kde is
apparently the only project of the several that's still putting the
developed code from the sponsored projects to use, tho further
development has slowed dramatically along with the flow of funding. I
wasn't involved personally and don't know the details, but I've read that
all the other projects dried up and blew away with the wind, once the
funding ran out.
So from that perspective, things aren't quite as bad as they might look.
KDE's actually doing pretty good as it's still using what they worked on,
while the others are all gone. And while development /has/ slowed down,
it hasn't stopped. KDE's still building on it, slowly. And if there is
ever more money made available, kde won't be starting from scratch this
time! It's one of the few projects that has something there to build on
already, and would thus be able to hit the ground running!
But while all that's nice, for me personally, I'm solidly in the computer-
technophile group, and don't expect that I'll find much use for the
semantic-desktop in any form, in the near future anyway. Maybe after
another half decade to decade of development... maybe not. But I don't
see myself interested in it near-term, for sure, especially now that I
know how much of a difference actually disabling that stuff at compile-
time can make!
But of course, disabling semantic-desktop in kde has a few knock-on
effects as well. Akonadi is closely enough related to semantic-desktop
that on gentoo, it requires USE=semantic-desktop. And all of kdepim,
including kmail, requires akonadi (directly or indirectly, thru kdepim-
common-libs). So no semantic-desktop support means no kdepim, no kmail,
etc. Additionally, there's a few of dolphin's features, etc, that are
semantic-desktop dependent. The file properties dialog has an
information tab, for instance, that's all semantic-desktop information.
So with semantic-desktop turned off, that tab is blank.
But it's still not worth USE=semantic-desktop. I'll keep my USE=-
semantic-desktop, thank-you-very-much. And if that becomes impossible
with kde, I'll likely end up switching to something other than kde. But
from what I've read, kde frameworks aka kde5, is being built with a goal
of more modularity, so if anything, building without semantic-desktop
should get easier, not more difficult. =:^)
--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman
___________________________________________________
This message is from the kde mailing list.
Account management: https://mail.kde.org/mailman/listinfo/kde.
Archives: http://lists.kde.org/.
More info: http://www.kde.org/faq.html.
