I find printer friendly version very convenient for archiving. However it seems to be restricted to a single message. I would like to save a complete thread in a printer friendly version. If not possible, could you add this option? Alejandro

Yes, I find that Maple 10 works fine under Debian Sarge. Furthermore, I have found that it also works fine under Knoppix 4.0. I have booted from the live CD, copied the Maple directory to the created file system and run Classic. Then, it seams feasible to make a live CD/DVD/USB including Maple and be able to use it anywhere, provided a suitable licence file is available.

Axel Thank you for showing this solution. I am not currently interested in this integral per se, but your lengthly solution makes me wonder about simplification techniques for expressions involving elliptic integrals. I would find hard to beleive that a more compact expression does not exist. A parallel issue being whether an explicitly real solution is available. Alejandro

I get different error messages for Windows and Linux, in both OSs with Maple 10.03: Win XP SP2: Error, (in evalf/int) NE_QUAD_ROUNDOFF_EXTRAPL: Round-off error is detected during extrapolation. Fedora Core 4: Error, (in evalf/int) NE_QUAD_MAX_SUBDIV: The maximum number of subdivisions has been reached: max_num_subint = 200 The last one is like yours. In either case these messages seem to be reproducible. I wonder what is the origin of this difference. I would not say that not changing the working precision on the fly is the Maple way. Rather, I understand Maple-like behavior as the possibility of setting control parameters by hand if the user wishes. But if an algorithm/heuristics exist that could provide good results by changing some parameters, I would welcome it also. Examples of implementations following this design exist elsewhere in the Maple system. Here two parameters seem to be under control of the user: digits and epsilon. Other parameters like number of subdivisions and timeout should, perhaps, be also available. Alejandro

Indeed, the integrand has an (analytically integrable) singularity at the upper end of the integration interval, as it behaves like 1/sqrt(vm-v) for v->vm: series(sech(v)-sech(vm), v=vm,2): map(normal@convert,%,exp); But this singularity occurs for any value of vm (you plot it for vm=1). So, I do not see any direct connection between this singularity and the errors that arise for small vm. At least, there should be something else that happens in that region. Yes, I have forgot to delay evaluation of N for the plot. Thanks. As said before, the default evalf/Int method returns the integral unevaluated, without the epsilon option, when vm is small. So, it seems unable to choose a suitable method to meet the default accuracy target (eps = 0.5 * 10^(1-digits) I guess). I wonder whether it also tries to change epsilon in case of failure. Alejandro

The error messages, when they appear (they do not appear for the default method, or method = _Dexp), may state the conditions that makes these algorithms stop, but they do not explain the reasons why they reach that condition. In particular why they occur for small vm and not for larger values. And because numerical and analytic methods are, in principle, unrelated I find curious that all these different numerical problems occur precisely where the integral is particularly smooth as a function of vm (it is almost a constant that can be calculated by elementary integration). It could be just chance, but without an clear explanation of the origin of these numerical errors, I cannot avoid thinking that there could be a relationship. And another strange issue occurs when trying to plot with your version: foo := proc(v) if sqrt(sech(v) - sech(VM)) = 0 then return Float(undefined) else return Re(cosh(v)/sqrt(sech(v) - sech(VM))) end if; end proc: N:=(vm,eps)->evalf(subs({VM=vm,EPS=eps},Int(eval(foo),0..VM,epsilon=EPS,method = _d01ajc))): I get negligible evaluation time for a point: t:=time(): N(.01,1.0e-7); time()-t; 2.221566424 0. But a plot like plot(N(v,1.0e-6),v=0.2..1); that takes less than a second without the epsilon option does not finish after several minutes (taking 50% of CPU resources). Alejandro

This link "here" points here: http://beta.mapleprimes.com/blog/will/news-discontinuation-of-maple-platforms instead of here: http://beta.mapleprimes.com/blog/will/news-maple-now-available-for-intel-based-macintoshes

Is this list closed? Or, is there chance for other Linux distributions like Debian or Fedora?

I had forgotten about it, but the command expandoff() could do the job. It has to be "defined" in a curious way: expand(expandoff()).

Certainly, the control of expand() is an issue. For that purpose I have used a call to: expand1:=proc(x) expand(x,op(select(type,indets(x),`^`))); end: which prevented some unwanted expansions. Better ideas are welcomed. I have experimented with frontend() also, but I decided not to use it as it was not very reliable: it has bugs not solved for years. I have reported some of them elsewhere.

Your code to handle "polynomials" with arbitrary powers contains some interesting ideas, though need to be worked out further to become a useful tool. For instance, an initial expansion step should be added to handle properly expressions like: a*t^(1+1/p)+h*t*(t^(1/p)+b*(t^(1/(2*p)))^2)+u^3+k*u^3; Long time ago (Maple V R4 and 5) I have written some code to collect and analize expressions with arbitrary powers, for the analysis of a class of ordinary differential equations. That code was a bit more lengthly, in part because some language facilities were not available then. Surely, if I happen to update that old code, I will take into account your ideas.

The last version of Maple where I get a correct plot background on this PC ("Classic" GUI, under Windows) is Maple V Release 5.1. Clearly something changed in the display of graphics after that. What?

I have set SELinux enabled again, booted, files were "relabeled", additionally I have set SELinux "enforced", hence back to the initial configuration (at least apparently), and Maple still works (currently Classic and command line in in fact, a firewall issue seems to block access to the kernel by Standard). It is not clear why, but this problem seems to be solved.

Yes, disabling SELinux does solve this problem. However, I would prefer a Maple specific security change. It seems that this is an issue for the selinux-list. OpenOffice is part of FC4 distribution, so there is no problem with it. I have not tried yet Mathematica or Matlab. Are the officially supported platforms for eg. Maple 10 exactly the same as those stated in the Maple 10 Installation and Licensing Guide - Appendix: System requirements? Beyond that, I would find useful to know on what linux distros Maple 10 or earlier work fine or not. I am not aware whether this information has been collected already. If not, I would suggest opening in this site a table Maple version vs linux distro say, where experiences with different combinations could be collected. Regards, Alejandro

With Maple 9.5, exactly the same installation and execution problems of Maple 10 occur. Curiously, Maple 9 GUI installation does work, but execution does not as with Maple 10. Maple 8 standard installation does not work: ======= [root@localhost cdrecorder]# ./installMapleLinuxSU Error occurred during initialization of VM Unable to load native library: /media/cdrecorder/Linux/Linux/resource/jre/lib/i386/libjava.so: symbol __libc_wait, version GLIBC_2.0 not defined in file libc.so.6 with link time reference ======= I have not found yet an option to make a console installation. And Maple 7 does install, but also fails in execution: ======= [root@localhost bin]# ./maple /usr/local/maple7/bin.IBM_INTEL_LINUX/cmaple: error while loading shared libraries: /usr/local/maple7/bin.IBM_INTEL_LINUX/libmclient.so: cannot restore segment prot after reloc: Permission denied ======= Though TestMaple also runs. Any idea?