OK, I'm just about ready to call time on this Linux experiment! I was
supposed to be evaluating this for possible deployment in a postgraduate GIS
suite. I have been using PC's for over 25 years (i.e. I am old enough to
have operated things via command line and done some machine code
programming!) and therefore I am not computer illiterate (at least compared
to our postgrads). I am not trying to upset those of you who like this sort
of challenge but this Linux thing frankly is not for general use. Few things
actually work - how can I recommend this when, for example, it talks to my
printers one minute and then decides they don't exist the next! The list of
"it nearly works" items seems endless.
This also seems to apply to GRASS. Does anyone know how to get Grass60
working on Fedora Core 5? I tried a reinstall and a clean installation of
just the tcl/tk8.4 libraries to no avail (segmentation fault in nviz -
assuming these libraries are the problem - how would I know?). I also tried
to download the 8.3 libraries as suggested by Hamish BUT then when I install
the grass-bin libraries the "clever" installer software decides it needs
tcl8.4 and downloads it anyway!
At present I am unlikely to recommend either Linux or GRASS for deployment.
It's a great idea but things need to be far more robust. I would however
VERY much like to see GRASS actually working. I am certain it has many
advantages over commercially available packages but it must be robust enough
to actually trust. I would be more than happy to admit that I have got this
wrong!
-----Original Message-----
From: Hamish [mailto:hamish_nospam@yahoo.com]
Sent: Wednesday, June 07, 2006 11:51 AM
To: Darrel Maddy
Subject: Re: [GRASS-user] Problem with Nviz : first post
Apologies if this has been covered elsewhere. Please also note that I
am new to Linux and GRASS (less than 4 days!!) - so it is safe to
assume that I know nothing!.
I have managed to get GRASS6 up and running on fedora core 5 (after
some "fun" installing tcl/tk libraries). When I enter Nviz and try to
display the elevation.dem from the Spearfish60 sample dataset the
visualisation appears for about 10 seconds and then disappears - no
error messages are reported to help sort the problem. The control
window is empty.
I may have confused the software with my tcl/tk installation antics -
I now have both tcl/tk 8.4 and 8.5 libaries in my /usr/lib directory
but only the 8.4 library is reported when I test the installation in
the command window. I do not know how to uninstall 8.5 (I could not
get GRASS to work with the 8.5 libraries). Is this the source of my
problem or am I completely missing the point?
Any help with what to try next would be greatly appreciated.
Many thanks
others using FC5 have reported similar problems recently.
TclTk 8.4 [threaded] only works with the latest versions of GRASS
6.1-cvs. For GRASS 6.0 use tcltk 8.3 or a distribution that uses an
untreaded version 8.4. (using 8.3 is easier)
We also have GRASS working fine with Fedora Core 4. We haven't yet upgraded
to 5. Given that Fedora is the open/cutting edge/kind of experimental
version of Red Hat, I've been reluctant to upgrade right away until others
with more time and experience have had a chance to work out the bugs.
If you want stable, I'd recommend Red Hat, FC4, Debian or others. Being on
the cutting edge can be fun and frustrating.
For different reasons, I also strongly recommend using 6.1 cvs version of
GRASS. Many things have been upgraded and fixed between 6.0 and 6.1. GRASS
tends to be very conservative in releasing stable versions. 6.1 may be more
reliable on current systems than 6.0 for this reason. It certainly works
better in most respects.
Michael
__________________________________________
Michael Barton, Professor of Anthropology
School of Human Evolution & Social Change
Center for Social Dynamics & Complexity
Arizona State University
From: Prof Darrel Maddy <darrel.maddy@ncl.ac.uk>
Date: Fri, 9 Jun 2006 12:19:39 +0100
To: <grassuser@grass.itc.it>
Subject: RE: [GRASS-user] Problem with Nviz : first post
OK, I'm just about ready to call time on this Linux experiment! I was
supposed to be evaluating this for possible deployment in a postgraduate GIS
suite. I have been using PC's for over 25 years (i.e. I am old enough to
have operated things via command line and done some machine code
programming!) and therefore I am not computer illiterate (at least compared
to our postgrads). I am not trying to upset those of you who like this sort
of challenge but this Linux thing frankly is not for general use. Few things
actually work - how can I recommend this when, for example, it talks to my
printers one minute and then decides they don't exist the next! The list of
"it nearly works" items seems endless.
This also seems to apply to GRASS. Does anyone know how to get Grass60
working on Fedora Core 5? I tried a reinstall and a clean installation of
just the tcl/tk8.4 libraries to no avail (segmentation fault in nviz -
assuming these libraries are the problem - how would I know?). I also tried
to download the 8.3 libraries as suggested by Hamish BUT then when I install
the grass-bin libraries the "clever" installer software decides it needs
tcl8.4 and downloads it anyway!
At present I am unlikely to recommend either Linux or GRASS for deployment.
It's a great idea but things need to be far more robust. I would however
VERY much like to see GRASS actually working. I am certain it has many
advantages over commercially available packages but it must be robust enough
to actually trust. I would be more than happy to admit that I have got this
wrong!
-----Original Message-----
From: Hamish [mailto:hamish_nospam@yahoo.com]
Sent: Wednesday, June 07, 2006 11:51 AM
To: Darrel Maddy
Subject: Re: [GRASS-user] Problem with Nviz : first post
Apologies if this has been covered elsewhere. Please also note that I
am new to Linux and GRASS (less than 4 days!!) - so it is safe to
assume that I know nothing!.
I have managed to get GRASS6 up and running on fedora core 5 (after
some "fun" installing tcl/tk libraries). When I enter Nviz and try to
display the elevation.dem from the Spearfish60 sample dataset the
visualisation appears for about 10 seconds and then disappears - no
error messages are reported to help sort the problem. The control
window is empty.
I may have confused the software with my tcl/tk installation antics -
I now have both tcl/tk 8.4 and 8.5 libaries in my /usr/lib directory
but only the 8.4 library is reported when I test the installation in
the command window. I do not know how to uninstall 8.5 (I could not
get GRASS to work with the 8.5 libraries). Is this the source of my
problem or am I completely missing the point?
Any help with what to try next would be greatly appreciated.
Many thanks
others using FC5 have reported similar problems recently.
TclTk 8.4 [threaded] only works with the latest versions of GRASS
6.1-cvs. For GRASS 6.0 use tcltk 8.3 or a distribution that uses an
untreaded version 8.4. (using 8.3 is easier)
OK, I'm just about ready to call time on this Linux experiment! The list of
"it nearly works" items seems endless.
Linux can be insanely frustrating at times. You shouldn't have to learn the intricacies of an OS to get your work done. On the other hand, once y'all get Linux set up, your students will be able to use GRASS without having to know those intricacies.
This also seems to apply to GRASS. Does anyone know how to get Grass60
working on Fedora Core 5? I tried a reinstall and a clean installation of
just the tcl/tk8.4 libraries to no avail (segmentation fault in nviz -
assuming these libraries are the problem - how would I know?). I also tried
to download the 8.3 libraries as suggested by Hamish BUT then when I install
the grass-bin libraries the "clever" installer software decides it needs
tcl8.4 and downloads it anyway!
I'm running GRASS on Gentoo and Mac OS X and haven't had the tcl problems you report, but I echo Michael's and Hamish's suggestion that you move to 6.1-cvs. It's not perfect, but it's much better and more feature-rich than 6.0. You will probably need to compile it yourself, which can also be daunting. I'm sure many of us (myself included) would be happy to post our configure scripts should you request it.
At present I am unlikely to recommend either Linux or GRASS for deployment.
It's a great idea but things need to be far more robust. I would however
VERY much like to see GRASS actually working. I am certain it has many
advantages over commercially available packages but it must be robust enough
to actually trust. I would be more than happy to admit that I have got this
wrong!
GRASS needs evangelists, but not self-sacrificial ones. If you try version 6.1-cvs and still have doubts, remember that if the GRASS install goes poorly, you'll bear some of the blame. Perhaps you would be better served to have a machine in your office running GRASS so you can learn at a slower pace, track the improvements, and be better prepared to recommend--and support--GRASS in the future.
Bottom line: if you think supporting Linux will be too troublesome, it's best to wait. I support Windows and Mac OS X in my center, and I don't push GRASS. People generally use Maptitude and Arc/GIS, and they can get their work done. Support-wise, it's also fairly easy to integrate a bunch of Windows PCs into a Windows-centric network and run an ESRI license server. It's a lot tougher to integrate Macs and Linux PCs into Active Directory. I would prefer things were less Microsoft-dominated, but I'd prefer a higher salary too...
Please let us know your decisions and how we can help.
Regards,
Jesse
--
Jesse Hamner
Coordinator, Social Sciences Computing Center
Emory University
+011 404.727.6929
I suggest installing Suse linux. I had no problem with the OS installation or installing grass (other than a missing library, and that was because I installed it on a 64 bit OS rather than 32 bit.] FWIW, this is the missing software I had to install:
-------------
the gsl-1.6-7.i586.rpm is missing. You can install it from: http://ftp.gwdg.de/pub/opensuse/distribution/SL-OSS-stable/inst-source/suse/i586/
or from your distribution CD-ROM.
regards,
Otto Dassau
--------------------------------------------------------
The grass documentation problems are another story.
Jesse Hamner wrote:
Professor Maddy,
OK, I'm just about ready to call time on this Linux experiment! The list of
"it nearly works" items seems endless.
Hang in there. if you managed to get grass and Linux installed and running in 3 days, then i’d say your off to a good start!
As someone recommended in previous post, stick with FC4 to take the road of least resistance, unless you want to be a pioneer in working through new release issues with the developers (which i’m sure is encouraged). I’ve installed GRASS from source on FC4, and would be glad to help, although i’m certainly not an expert. If you work with the community in posting errors and resolving problems, the rewards will be plentiful and favor both parties involved. It’s one of the ways open source software gains strength.
Your situation is special and very important: having the ability to teach GIS on an open source platform to students who will go forth and use GRASS in their careers. They will in turn help use, and possibly even contribute to, the evolution of GRASS. Not everyone has the capability to show the masses this amazing alternative to commercial GIS and OS.
“I am certain it has many advantages over commercially available packages but it must be robust enough to actually trust.”
the robustness (reliability) you seek is there, as well as the advantages. But first the system (Linux & GRASS) must be running correctly for that attribute to be applicable. I have to use ESRI software (due to situational circumstances), that is $14K USD retail last i was told (that price doesnt include the raster analysis module). I have been using it for over 10 years, and lived through the abandonment of the UNIX world command line (around 2000), and software re-write to windows @rcMap, which in my opinion is the antonym of robust; so much to the point that i use the command line workstation version to come as close to robustness as i can in a windows environment for data creation and modification. @rcMap is the reason I sought out other GIS software, and I couldnt be more pleased to have entered the world of GRASS and experience GIS the way it was meant to be.
Please consider, however, that if you're ready to invest, say 50% of what you
invest in your Mircosoft+ESRI+whatever licences, you'll get a system that is
more reliable, stable, and advanced, than others. Of course, if you don't
invest (either time to learn, or money to have support for the
installation/configuration/feature enhancement/bug fixing problems you may
have), you may end up not getting very much.
Don't assume free software is something not worth/necessary to invest in. It's
rather obvious to many of us, but sometimes is good to repeat: think of free
as free thinking, not as free beer.
All the best.
pc
At 23:52, sabato 10 giugno 2006, Jesse Hamner has probably written:
Professor Maddy,
...
Bottom line: if you think supporting Linux will be too troublesome, it's
best to wait. I support Windows and Mac OS X in my center, and I don't
push GRASS. People generally use Maptitude and Arc/GIS, and they can get
their work done. Support-wise, it's also fairly easy to integrate a
bunch of Windows PCs into a Windows-centric network and run an ESRI
license server. It's a lot tougher to integrate Macs and Linux PCs into
Active Directory. I would prefer things were less Microsoft-dominated,
but I'd prefer a higher salary too...
re. linux's difficulties, "It's a steep learning curve, but there are
wonderful views from the top."
This also seems to apply to GRASS. Does anyone know how to get Grass60
working on Fedora Core 5? I tried a reinstall and a clean installation
of just the tcl/tk8.4 libraries to no avail (segmentation fault in
nviz - assuming these libraries are the problem - how would I know?).
I also tried to download the 8.3 libraries as suggested by Hamish BUT
then when I install the grass-bin libraries the "clever" installer
software decides it needs tcl8.4 and downloads it anyway!
GRASS 6.0.2's NVIZ is not compatible with a TclTk 8.4 or newer built
with threading support.
This bug is fixed now and NVIZ works with TclTk 8.4 in the development
version. A future 6.0.3 release can/will have it working as well.
Sorry I can't help any with FC5, just note that it is very new so few of
us will have any experience with it. Why go through the inconvenience of
building GRASS yourself when you can use the pre-built version for FC5
from Intevation GmbH? (including NVIZ bug I think) http://ftp.gwdg.de/pub/misc/freegis/intevation/freegis/fedora/5/
If you are unhappy with linux (really FC5 !!) try the MacOSX version
which works "out of the box, no assembly required". (I only have recent
experince with Lorenzo Moretti's version which is very well done)
There are linux versions like Debian (or Ubuntu) where there is a
"DebianGIS" QA team who have already worked through all these problems
for you and a 100% working install is as easy as typing
"apt-get install grass" from the command line. Ubuntu is a Debian
derivative which is better suited for new users.
the robustness (reliability) you seek is there, as well as the
advantages. But first the system (Linux & GRASS) must be running
correctly for that attribute to be applicable. I have to use ESRI
software (due to situational circumstances), that is $14K USD retail
last i was told (that price doesnt include the raster analysis
module). I have been using it for over 10 years, and lived through
the abandonment of the UNIX world command line (around 2000), and
software re-write to windows @rcMap, which in my opinion is the
antonym of robust; so much to the point that i use the command line
workstation version to come as close to robustness as i can in a
windows environment for data creation and modification. @rcMap is the
reason I sought out other GIS software, and I couldnt be more pleased
to have entered the world of GRASS and experience GIS the way it was
meant to be.
Entertaining experiment I read about recently in an interview with
ESRI's CEO: (comment by the author, not CEO!)
Go to ESRI's website and try the "driving directions to HQ" link.
(ArcMap+ArcIMS)
Then go to google maps (build in part with familiar Open Source tools)
and try with the same address.
Apparently Google's map is many many times nicer. I never got to compare
though as the ESRI map was forever stalled and never loaded. I have had
a similarly poor (mostly just very very slow) experience with *every
single* ArcIMS server I have ever used.
People still don't trust MapServer but will pay many thousands of
dollars for a branded product that hardly works.. sigh.
Another thought experiment: how many hours (years) of Linux tech support
can you buy from a local computer science undergraduate student with the
funds it takes to buy a single software license of a commercial GIS?
What else could you use that money for? How many student scholarships?
Hamish
ps - (repeated) if you don't know linux, try the mac version
