# Lurupa Documentation

### 1.0

Lurupa is a tool for computing rigorous error bounds in linear programming.

This documentation consists of four parts.

• Introduction
We will start with the question why at all do we need such tools. What can go wrong when we solve a linear program? What can a verification tool do for us in these cases?
• Theory
After that we look at the mathematics behind the computations done in Lurupa. What happens in the algorithms and how can we make sure that the computed bounds are rigorously, including rounding errors.
• Usage
The usage of Lurupa has to be divided into two parts.
• Stand-alone usage
In this section we will investigate the usage as a stand-alone command line tool.
• API usage
Here we look at the usage through the API.
Finally the terms under which you are allowed to use Lurupa.

## Introduction

The usefulness of verification tools for linear programming can be demonstrated with simple examples. Let's consider the following simple linear program

While the infeasibilty of this problem can be easily seen, many linear programming solvers like CPLEX 9.000, lp_solve 5.5, and MATLAB R14SP3's linprog will regard this problem as feasible. With the default accuracy they will happily return with an optimal solution not even displaying a warning.

A similar construction for an unbounded problem,

fools CPLEX and lp_solve into returning 0 as the optimal solution for the variables and thus as the optimal function value.

Given, these examples are very simple. They are just used to illustrate that even in these dimensions problems may arrise. Adjusting the accuracy of the solvers would enable them to return the correct results. One may ask if 0.999 is different from 1? Is 0.999999 different? What about 0.999999999999? If not the solvers are correct. The values used above, however, are recognized by the solvers as being different from 1.

Going to higher dimensions the problem becomes more subtle. The dependencies between the constraints cannot be seen easily. Let us look at a problem from the Netlib lp library [7] , a free collection of insdustrial and academical linear programs. With 56 constraints and 97 variables adlittle is one of its smaller members. While being in fact feasible, adlittle suffers from ill--posedness. Perturbing the right hand side of the equality constraints by subtracting a tiny multiple of the 96th column of the equation matrix renders the linear program infeasible. Running this problem through CPLEX and lp_solve does again return a solution without any warnings.

Now we will have a look at what Lurupa returns when confronted with these problems. Using lp_solve to compute approximate solutions as a start off, Lurupa returns for the primal infeasible problem the optimal value interval . While the lower bound does not help us much here, the upper bound tells us, that the problem is at least very close to infeasibility as Lurupa could not find verified feasible points. What this means is explained in the next section. Computing bounds for the second, unbounded problem Lurupa returns the interval . Here we can pinpoint the unboundedness in the infinite lower bound. Using the final example as input, Lurupa computes the bounds (we need the resetbas option of the solver module in this case to deal with the bad numerics). Again the upper bound indicates the problem being at least very close to infeasibility.

On the other hand let us have a look at what Lurupa does when confronted with a feasible lp. Taking adlittle again, now with the original coefficients, Lurupa returns the interval for the optimal value. This does not only give a verified interval containing the optimal value, but also verifies that feasible solutions in fact exist. This problem cannot be infeasible or unbounded, as the lower and upper bounds prove the existence of dual and primal feasible points, respectively.

Results of Lurupa being applied to the whole selection of Netlib problems is contained in [5] .

## Theory

The idea behind the computations done in Lurupa is to use a combination of iteratively perturbing the linear program and using interval arithmetic to verify feasible points of the original linear program. Once such a point is verified a rigorous bound on the optimal value can be derived. The details can be found in the work of Jansson [3] . Here only an overview is given.

### Interval Arithmetic

Before we look at the actual algorithms, we need a tool to ensure computations to be correct even in the presence of rounding errors. This can be achieved with interval arithmetic. The basic idea is to not represent the value of an expression by a number but by an interval that is guaranteed to contain the exact value.

Let us have a look at a simple example. We use a decimal floating point arithmetic with only 2 digits mantissa and compute

Here means the floating point evaluation of . The 0.1 is completely annihilated by rounding the result to 2 digits. Repeating this using an interval arithmetic gives us the following

We arrive at a guaranteed enclosure, 10.1 is certainly between 10 and 11.

But how do we get there? The equation gives a hint how to arrive at this result. The interval enclosure of the expression is formed by the lower bound and the upper bound . These are computed using directed rounding towards minus infinity and plus infinity, respectively.

Having arrived at intervals we now have to continue our computations using these. The objective stays just the same. Compute an interval that is guaranteed to contain the correct result using directed rounding. As we do not know which number in the intervals is the correct one, we have to take all into account. If we have to divide the result of the previous example by , this means

The lower bound is the result of rounding 10 over 3 downwards, and the upper bound stems from rounding 11 over 2 upwards.

For a more thorough introduction to interval arithmetic have a look at Kearfott [4] and Hayes [2] and the references therein. The interval computations homepage [6] is also a good starting point.

### Rigorous lower and upper bounds

Looking at a linear program in standard form

we use a standard linear programming solver to compute an approximate solution . We make no assumptions about the accuracy of the computed approximations. A good approximation, however, usually results in a sharper enclosure of the optimal value.

The next step is to enforce the simple bounds . We achieve this by setting the components of that violate their corresponding bound to the violated bound. If or , we set or , respectively.

Now we use an interval linear system solver to compute a rigorous enclosure of the solution set of . This step transforms the point vector into an interval vector , which is verified to contain at least one point satisfying the equations.

Finally we just have to check if the inequalities and the simple bounds are satisfied by and thus by all points contained in the interval vector. If this is the case, contains at least one point that satisfies all constraints. Taking the maximum of the objective function over the box , that is over the objective value of all points in the interval vector, we arrive at an objective value that is verified to be larger than the one of a feasible point. This in turn is larger than the optimal value, which leaves us with an upper bound on the optimal value.

If does not satisfy the constraints, the iteration starts. We perturb the linear program by lowering the right hand sides of the inequalities and tightening the simple bounds . Using this perturbed linear program as an input to the linear programming solver, we compute a new approximate solution . With the above procedure we try again to derive an interval vector verified to contain a feasible point of the original linear program.

If the linear programming solver does not return an approximate solution in any stage of the iteration, due to numerical failure for example, the only upper bound we can derive is plus infinity. The same holds if the interval linear system solver cannot compute an enclosure of the solution set of the equations. As the equations do not change during the iteration, this can be ruled out after the first iteration.

The lower bound on the optimal value can be derived in a similar way from the dual linear program

Here, however, we can make use of the special structure of the dual, thus reducing the computational complexity. The details are contained in [3] .

## Usage

Lurupa can be used in two ways.

In each case an lp solver specific module is needed to translate the generic calls from Lurupa into calls to the solver API and to transfer data between Lurupa and the solver.

### Stand-alone usage

The command line client lurupa is controlled through various options. Mandatory are the options to select an lp and a solver module.
-lp file
This option specifies the file containing the lp to be processed. The file has to be in a format that can be handled by the solver module.
-sm file
This option specifies the solver module and thus the lp solver that is used to compute the approximate solutions.

The most basic call, using a solver module Module, assumend to be located in the local directory, would thus be

 lurupa -sm Module -lp lp.mps

Here we assumed the module knows how to parse lp.mps. This call, however, would just call the solver through the module and solve the lp. To compute the bounds we have to specify which bounds we want. This is done via the bound parameters
-lb
This option specifies, that the lower bound for the optimal value shall be computed.
-ub
This option specifies, that the upper bound for the optimal value shall be computed.
Computing both bounds for lp.mps is therefor done with
 lurupa -sm Module -lp lp.mps -lb -ub


It's important to note that while the computed bounds are rigorous the command line client does not provide for rigorous printing of the bounds. This means that the values printed for the bounds are rounded to display precision.

Several more options are available to alter Lurupa's behaviour. These can be divided into options affecting the input, the algorithm, and the output.

Options affecting the input

-i d
This option specifies, that the cost vector and the constraint matrices and right hand sides are inflated to intervals with the given midpoint and a radius of d. The parameters get multiplied by the interval .

Options affecting the algorithm

-alpha d
This option changes the algorithm parameter to d. For details on the parameter refer to [3] .
-eta d
This option changes the algorithm parameter to d. For details on the parameter refer to [3] .
-inflate
This option tells Lurupa to inflate perturbed lps in case of infeasibility. Due to the exponential deflation used in the algorithm the gap between lps that can not be verified and lps that are considered infeasible by the solver might be missed. If this option is specified Lurupa tries to inflate the feasible region of perturbed, infeasible lps to reach this gap. Using this option increases the number of algorithm iterations.

Options affecting the output

-csv file
This option tells Lurupa to append a line containing the comma seperated results to file. If not present, the extension .csv is appended to the filename.
-latex file
This option tells Lurupa to append a line contating the results to file, formatted to be used in a LaTeX table. If not present, the extension .tex is appended to the filename.
-t
This option causes all messages printed by Lurupa to be prepended with the current time.
-vn
This option selects the verbosity level of the messages printed by Lurupa. The verbosity increases with n. Possible values are:
-v0: No messages
Print no messages from Lurupa. The command line client just prints the final results.
-v1: Errors
Add error messages from Lurupa to -v0.
-v2: Warnings
-v3: Brief
Add brief program flow messages like the iteration number and whether we are currently approximately solving a perturbed problem or verifying the feasibility of a solution.
-v4: Normal
Add more program flow messages like whether we are perturbing or approximately solving a perturbed lp, and intermediate results like the approximate optimal value of perturbed lps.
-v5: Verbose
Add detailed information like changes of the deflation parameters.
The default verbosity level is -v2.
-write_vm style
This option tells Lurupa to save intermediate vectors and matrices to disk. The format depends on style. Specifying octave stores all these in readable by Octave, this means all in one file with meta information about the dimension and the name of the vector or matrix. The file is named after the lp. Specifying matlab stores these in one file per vector or matrix. These files are put into a directory named after the model.

Remain the options that print informational messages.

-h, -?, --help
These options tell Lurupa to print usage information about the command line client containing the supported parameters along with all parameters supported by the solver module if specified.
-V
This option tells Lurupa to print version information about the command line client, the core, and the solver module if specified. This also contains values like the set algorithm parameters.
-Vb
This option tells Lurupa to print a brief version of the information printed by -V.

As mentioned the solver module might support additional options. These are specific to the solver module. They have the common prefix -sm, with additional parameters appended with commas. A typical option would be -sm,timeout,3600, which might tell the solver to time out after 3600 seconds.

#### lp_solve solver modules

There are solver modules for different versions of lp_solve [1] . Development is done on the latest version of lp_solve with features backported as appropriate. Thus the modules support different options

lp_solve 3.2 No further options

lp_solve 4.0.1.0 and 4.0.1.11

-sm,timeout,i
This option sets the timeout in seconds to i. Setting this calls lp_solve's set_timeout, causing lp_solve to stop solving an lp if it takes longer than i seconds.
-sm,vn
This option selects the verbosity level of the messages printed by lp_solve. The verbosity increases with n. Possible values are (taken from the lp_solve documentation):
v0: NEUTRAL
Only some specific debug messages in debug print routines are reported.
v1: CRITICAL
Only critical messages are reported. Hard errors like instability, out of memory, ...
v2: SEVERE
Only severe messages are reported. Errors.
v3: IMPORTANT
Only important messages are reported. Warnings and Errors.
v4: NORMAL
Normal messages are reported.
v5: DETAILED
Detailed messages are reported. Like model size, continuing B&B improvements, ...
v6: FULL
All messages are reported. Useful for debugging purposes and small models.
The default verbosity level is v3.

lp_solve 5.5

-sm,resetbas
This option causes the solver module to re-solve lps in the case of numerical failure with an all-slack basis. While this increases the runtime it may help with numerically difficult lps.
-sm,s
This option enables the default scaling of lp_solve (Numerical range-based scaling).
-sm,timeout,i
This option sets the timeout in seconds to i. Setting this calls lp_solve's set_timeout, causing lp_solve to stop solving an lp if it takes longer than i seconds.
-sm,vn
This option selects the verbosity level of the messages printed by lp_solve. The verbosity increases with n. Possible values are (taken from the lp_solve documentation):
v0: NEUTRAL
Only some specific debug messages in debug print routines are reported.
v1: CRITICAL
Only critical messages are reported. Hard errors like instability, out of memory, ...
v2: SEVERE
Only severe messages are reported. Errors.
v3: IMPORTANT
Only important messages are reported. Warnings and Errors.
v4: NORMAL
Normal messages are reported.
v5: DETAILED
Detailed messages are reported. Like model size, continuing B&B improvements, ...
v6: FULL
All messages are reported. Useful for debugging purposes and small models.
The default verbosity level is v3.
-sm,wmps
This option causes the solver module to save the perturbed lps in mps format. The filename is constructed of the lp name and the current iteration number.

### API usage

Using Lurupa through the API means greater flexibility and allows to embed Lurupa in a larger framework.

C++

To access the API we need to include the two header files globals.h and Lurupa.h. Now we need to create an instance of the Lurupa class. All interaction with Lurupa is done via this object. As in the stand-alone case we need a solver module, which selects the lp solver to be used. Selecting the lp to be solved, we arrive at the following code that will appear more a less like this in every API usage of Lurupa.

 #include <lurupa/globals.h>
#include <lurupa/Lurupa.h>

[...]

Lurupa lurupa;
bool module_set = lurupa.set_module("path/to/module");

The second parameter to read_lp reads the lp as it is and does not inflate it to an interval problem.

Now we can compute an approximate solution and after that the rigorous bounds on the optimal value.

 double optimal_value;
Solver_status status;
bool lp_solved = lurupa.solve_lp(optimal_value, status);

double lower_bound, upper_bound;
int lower_iterations, upper_iterations;
Bound_status st_lower = lurupa.lower_bound(lower_bound, lower_iterations);
Bound_status st_upper = lurupa.upper_bound(upper_bound, upper_iterations);


This is all that is necessary to compute rigorous bounds on the optimal value. For details on how to set the algorithm parameters, how to change output and the like refer to the reference of the class Lurupa.

C

There is a C wrapper for the library, which maps all calls to an internal Lurupa object. To use it we have to include the cLurupa.h header. The names of the routines are chosen after the names of the corresponding member routines of Lurupa with a prefix of clu. The previous C++ example looks like the following using the C wrapper.

 #include <lurupa/cLurupa.h>

[...]

double optimal_value, lower_bound, upper_bound;
BOUND_STATUS st_lower, st_upper;
int lower_iterations, upper_iterations;

clu_init();
module_set = clu_set_module("path/to/module");
lp_solved = clu_solve_lp(&optimal_value, &status);
st_lower = clu_lower_bound(&lower_bound, &lower_iterations);
st_upper = clu_upper_bound(&upper_bound, &upper_iterations);

For a complete list of the wrapper functions refer to the reference of the cLurupa.h header.

Copyright (C) 2006 by Christian Keil

Lurupa is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

		  GNU LESSER GENERAL PUBLIC LICENSE
Version 2.1, February 1999

Copyright (C) 1991, 1999 Free Software Foundation, Inc.
51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.

[This is the first released version of the Lesser GPL.  It also counts
as the successor of the GNU Library Public License, version 2, hence
the version number 2.1.]

Preamble

The licenses for most software are designed to take away your
freedom to share and change it.  By contrast, the GNU General Public
free software--to make sure the software is free for all its users.

specially designated software packages--typically libraries--of the
Free Software Foundation and other authors who decide to use it.  You
can use it too, but we suggest you first think carefully about whether
strategy to use in any particular case, based on the explanations below.

When we speak of free software, we are referring to freedom of use,
not price.  Our General Public Licenses are designed to make sure that
you have the freedom to distribute copies of free software (and charge
for this service if you wish); that you receive source code or can get
it if you want it; that you can change the software and use pieces of
it in new free programs; and that you are informed that you can do
these things.

To protect your rights, we need to make restrictions that forbid
distributors to deny you these rights or to ask you to surrender these
rights.  These restrictions translate to certain responsibilities for
you if you distribute copies of the library or if you modify it.

For example, if you distribute copies of the library, whether gratis
or for a fee, you must give the recipients all the rights that we gave
you.  You must make sure that they, too, receive or can get the source
code.  If you link other code with the library, you must provide
complete object files to the recipients, so that they can relink them
with the library after making changes to the library and recompiling
it.  And you must show them these terms so they know their rights.

We protect your rights with a two-step method: (1) we copyright the
library, and (2) we offer you this license, which gives you legal
permission to copy, distribute and/or modify the library.

To protect each distributor, we want to make it very clear that
there is no warranty for the free library.  Also, if the library is
modified by someone else and passed on, the recipients should know
that what they have is not the original version, so that the original
author's reputation will not be affected by problems that might be
introduced by others.

Finally, software patents pose a constant threat to the existence of
any free program.  We wish to make sure that a company cannot
effectively restrict the users of a free program by obtaining a
restrictive license from a patent holder.  Therefore, we insist that
any patent license obtained for a version of the library must be
consistent with the full freedom of use specified in this license.

Most GNU software, including some libraries, is covered by the
General Public License, applies to certain designated libraries, and
is quite different from the ordinary General Public License.  We use
libraries into non-free programs.

When a program is linked with a library, whether statically or using
a shared library, the combination of the two is legally speaking a
combined work, a derivative of the original library.  The ordinary
entire combination fits its criteria of freedom.  The Lesser General
the library.

does Less to protect the user's freedom than the ordinary General
Public License.  It also provides other free software developers Less
are the reason we use the ordinary General Public License for many
special circumstances.

For example, on rare occasions, there may be a special need to
encourage the widest possible use of a certain library, so that it becomes
a de-facto standard.  To achieve this, non-free programs must be
allowed to use the library.  A more frequent case is that a free
library does the same job as widely used non-free libraries.  In this
case, there is little to gain by limiting the free library to free
software only, so we use the Lesser General Public License.

In other cases, permission to use a particular library in non-free
programs enables a greater number of people to use a large body of
free software.  For example, permission to use the GNU C Library in
non-free programs enables many more people to use the whole GNU
operating system, as well as its variant, the GNU/Linux operating
system.

Although the Lesser General Public License is Less protective of the
users' freedom, it does ensure that the user of a program that is
linked with the Library has the freedom and the wherewithal to run
that program using a modified version of the Library.

The precise terms and conditions for copying, distribution and
modification follow.  Pay close attention to the difference between a
"work based on the library" and a "work that uses the library".  The
former contains code derived from the library, whereas the latter must
be combined with the library in order to run.

TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION

0. This License Agreement applies to any software library or other
program which contains a notice placed by the copyright holder or
other authorized party saying it may be distributed under the terms of

A "library" means a collection of software functions and/or data
prepared so as to be conveniently linked with application programs
(which use some of those functions and data) to form executables.

The "Library", below, refers to any such software library or work
which has been distributed under these terms.  A "work based on the
Library" means either the Library or any derivative work under
copyright law: that is to say, a work containing the Library or a
portion of it, either verbatim or with modifications and/or translated
straightforwardly into another language.  (Hereinafter, translation is
included without limitation in the term "modification".)

"Source code" for a work means the preferred form of the work for
making modifications to it.  For a library, complete source code means
all the source code for all modules it contains, plus any associated
interface definition files, plus the scripts used to control compilation
and installation of the library.

Activities other than copying, distribution and modification are not
covered by this License; they are outside its scope.  The act of
running a program using the Library is not restricted, and output from
such a program is covered only if its contents constitute a work based
on the Library (independent of the use of the Library in a tool for
writing it).  Whether that is true depends on what the Library does
and what the program that uses the Library does.

1. You may copy and distribute verbatim copies of the Library's
complete source code as you receive it, in any medium, provided that
you conspicuously and appropriately publish on each copy an
appropriate copyright notice and disclaimer of warranty; keep intact
all the notices that refer to this License and to the absence of any
warranty; and distribute a copy of this License along with the
Library.

You may charge a fee for the physical act of transferring a copy,
and you may at your option offer warranty protection in exchange for a
fee.

2. You may modify your copy or copies of the Library or any portion
of it, thus forming a work based on the Library, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:

a) The modified work must itself be a software library.

b) You must cause the files modified to carry prominent notices
stating that you changed the files and the date of any change.

c) You must cause the whole of the work to be licensed at no
charge to all third parties under the terms of this License.

d) If a facility in the modified Library refers to a function or a
table of data to be supplied by an application program that uses
the facility, other than as an argument passed when the facility
is invoked, then you must make a good faith effort to ensure that,
in the event an application does not supply such function or
table, the facility still operates, and performs whatever part of
its purpose remains meaningful.

(For example, a function in a library to compute square roots has
a purpose that is entirely well-defined independent of the
application.  Therefore, Subsection 2d requires that any
application-supplied function or table used by this function must
be optional: if the application does not supply it, the square
root function must still compute square roots.)

These requirements apply to the modified work as a whole.  If
identifiable sections of that work are not derived from the Library,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works.  But when you
distribute the same sections as part of a whole which is a work based
on the Library, the distribution of the whole must be on the terms of
entire whole, and thus to each and every part regardless of who wrote
it.

Thus, it is not the intent of this section to claim rights or contest
your rights to work written entirely by you; rather, the intent is to
exercise the right to control the distribution of derivative or
collective works based on the Library.

In addition, mere aggregation of another work not based on the Library
with the Library (or with a work based on the Library) on a volume of
a storage or distribution medium does not bring the other work under

3. You may opt to apply the terms of the ordinary GNU General Public
this, you must alter all the notices that refer to this License, so
that they refer to the ordinary GNU General Public License, version 2,
ordinary GNU General Public License has appeared, then you can specify
that version instead if you wish.)  Do not make any other change in
these notices.

Once this change is made in a given copy, it is irreversible for
that copy, so the ordinary GNU General Public License applies to all
subsequent copies and derivative works made from that copy.

This option is useful when you wish to copy part of the code of
the Library into a program that is not a library.

4. You may copy and distribute the Library (or a portion or
derivative of it, under Section 2) in object code or executable form
under the terms of Sections 1 and 2 above provided that you accompany
it with the complete corresponding machine-readable source code, which
must be distributed under the terms of Sections 1 and 2 above on a
medium customarily used for software interchange.

source code from the same place satisfies the requirement to
distribute the source code, even though third parties are not
compelled to copy the source along with the object code.

5. A program that contains no derivative of any portion of the
Library, but is designed to work with the Library by being compiled or
linked with it, is called a "work that uses the Library".  Such a
work, in isolation, is not a derivative work of the Library, and
therefore falls outside the scope of this License.

However, linking a "work that uses the Library" with the Library
creates an executable that is a derivative of the Library (because it
contains portions of the Library), rather than a "work that uses the
library".  The executable is therefore covered by this License.
Section 6 states terms for distribution of such executables.

When a "work that uses the Library" uses material from a header file
that is part of the Library, the object code for the work may be a
derivative work of the Library even though the source code is not.
Whether this is true is especially significant if the work can be
linked without the Library, or if the work is itself a library.  The
threshold for this to be true is not precisely defined by law.

If such an object file uses only numerical parameters, data
structure layouts and accessors, and small macros and small inline
functions (ten lines or less in length), then the use of the object
file is unrestricted, regardless of whether it is legally a derivative
work.  (Executables containing this object code plus portions of the
Library will still fall under Section 6.)

Otherwise, if the work is a derivative of the Library, you may
distribute the object code for the work under the terms of Section 6.
Any executables containing that work also fall under Section 6,
whether or not they are linked directly with the Library itself.

6. As an exception to the Sections above, you may also combine or
link a "work that uses the Library" with the Library to produce a
work containing portions of the Library, and distribute that work
under terms of your choice, provided that the terms permit
modification of the work for the customer's own use and reverse
engineering for debugging such modifications.

You must give prominent notice with each copy of the work that the
Library is used in it and that the Library and its use are covered by
this License.  You must supply a copy of this License.  If the work
during execution displays copyright notices, you must include the
copyright notice for the Library among them, as well as a reference
directing the user to the copy of this License.  Also, you must do one
of these things:

a) Accompany the work with the complete corresponding
machine-readable source code for the Library including whatever
Sections 1 and 2 above); and, if the work is an executable linked
with the Library, with the complete machine-readable "work that
uses the Library", as object code and/or source code, so that the
user can modify the Library and then relink to produce a modified
executable containing the modified Library.  (It is understood
that the user who changes the contents of definitions files in the
Library will not necessarily be able to recompile the application
to use the modified definitions.)

b) Use a suitable shared library mechanism for linking with the
Library.  A suitable mechanism is one that (1) uses at run time a
copy of the library already present on the user's computer system,
rather than copying library functions into the executable, and (2)
will operate properly with a modified version of the library, if
the user installs one, as long as the modified version is
interface-compatible with the version that the work was made with.

c) Accompany the work with a written offer, valid for at
least three years, to give the same user the materials
specified in Subsection 6a, above, for a charge no more
than the cost of performing this distribution.

specified materials from the same place.

e) Verify that the user has already received a copy of these
materials or that you have already sent this user a copy.

For an executable, the required form of the "work that uses the
Library" must include any data and utility programs needed for
reproducing the executable from it.  However, as a special exception,
the materials to be distributed need not include anything that is
normally distributed (in either source or binary form) with the major
components (compiler, kernel, and so on) of the operating system on
which the executable runs, unless that component itself accompanies
the executable.

restrictions of other proprietary libraries that do not normally
accompany the operating system.  Such a contradiction means you cannot
use both them and the Library together in an executable that you
distribute.

7. You may place library facilities that are a work based on the
Library side-by-side in a single library together with other library
facilities not covered by this License, and distribute such a combined
library, provided that the separate distribution of the work based on
the Library and of the other library facilities is otherwise
permitted, and provided that you do these two things:

a) Accompany the combined library with a copy of the same work
based on the Library, uncombined with any other library
Sections above.

b) Give prominent notice with the combined library of the fact
that part of it is a work based on the Library, and explaining
where to find the accompanying uncombined form of the same work.

the Library except as expressly provided under this License.  Any
distribute the Library is void, and will automatically terminate your
or rights, from you under this License will not have their licenses
terminated so long as such parties remain in full compliance.

9. You are not required to accept this License, since you have not
signed it.  However, nothing else grants you permission to modify or
distribute the Library or its derivative works.  These actions are
prohibited by law if you do not accept this License.  Therefore, by
modifying or distributing the Library (or any work based on the
Library), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Library or works based on it.

10. Each time you redistribute the Library (or any work based on the
original licensor to copy, distribute, link with or modify the Library
subject to these terms and conditions.  You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties with

11. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
excuse you from the conditions of this License.  If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Library at all.  For example, if a patent
license would not permit royalty-free redistribution of the Library by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Library.

If any portion of this section is held invalid or unenforceable under any
particular circumstance, the balance of the section is intended to apply,
and the section as a whole is intended to apply in other circumstances.

It is not the purpose of this section to induce you to infringe any
patents or other property right claims or to contest validity of any
such claims; this section has the sole purpose of protecting the
integrity of the free software distribution system which is
generous contributions to the wide range of software distributed
through that system in reliance on consistent application of that
system; it is up to the author/donor to decide if he or she is willing
to distribute software through any other system and a licensee cannot
impose that choice.

This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.

12. If the distribution and/or use of the Library is restricted in
certain countries either by patents or by copyrighted interfaces, the
an explicit geographical distribution limitation excluding those countries,
so that distribution is permitted only in or among countries not thus
excluded.  In such case, this License incorporates the limitation as if
written in the body of this License.

13. The Free Software Foundation may publish revised and/or new
versions of the Lesser General Public License from time to time.
Such new versions will be similar in spirit to the present version,
but may differ in detail to address new problems or concerns.

Each version is given a distinguishing version number.  If the Library
specifies a version number of this License which applies to it and
"any later version", you have the option of following the terms and
conditions either of that version or of any later version published by
the Free Software Foundation.  If the Library does not specify a
the Free Software Foundation.

14. If you wish to incorporate parts of the Library into other free
programs whose distribution conditions are incompatible with these,
write to the author to ask for permission.  For software which is
copyrighted by the Free Software Foundation, write to the Free
Software Foundation; we sometimes make exceptions for this.  Our
decision will be guided by the two goals of preserving the free status
of all derivatives of our free software and of promoting the sharing
and reuse of software generally.

NO WARRANTY

15. BECAUSE THE LIBRARY IS LICENSED FREE OF CHARGE, THERE IS NO
WARRANTY FOR THE LIBRARY, TO THE EXTENT PERMITTED BY APPLICABLE LAW.
EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR
OTHER PARTIES PROVIDE THE LIBRARY "AS IS" WITHOUT WARRANTY OF ANY
KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE
LIBRARY IS WITH YOU.  SHOULD THE LIBRARY PROVE DEFECTIVE, YOU ASSUME
THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.

16. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY
AND/OR REDISTRIBUTE THE LIBRARY AS PERMITTED ABOVE, BE LIABLE TO YOU
FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE
LIBRARY (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
FAILURE OF THE LIBRARY TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGES.

END OF TERMS AND CONDITIONS

How to Apply These Terms to Your New Libraries

If you develop a new library, and you want it to be of the greatest
possible use to the public, we recommend making it free software that
everyone can redistribute and change.  You can do so by permitting
redistribution under these terms (or, alternatively, under the terms of the

To apply these terms, attach the following notices to the library.  It is
safest to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least the
"copyright" line and a pointer to where the full notice is found.

<one line to give the library's name and a brief idea of what it does.>
Copyright (C) <year>  <name of author>

This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
version 2.1 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA

Also add information on how to contact you by electronic and paper mail.

You should also get your employer (if you work as a programmer) or your
school, if any, to sign a "copyright disclaimer" for the library, if
necessary.  Here is a sample; alter the names:

Yoyodyne, Inc., hereby disclaims all copyright interest in the
library Frob' (a library for tweaking knobs) written by James Random Hacker.

<signature of Ty Coon>, 1 April 1990
Ty Coon, President of Vice

That's all there is to it!

`

Generated on Thu Jun 26 18:08:52 2008 for Lurupa by  1.5.6