1 <!doctype linuxdoc system>
4 $Id: egate.sgml,v 1.9 1996/04/25 10:21:53 adam Exp $
8 <title>Email/Z39.50 gateway guide
9 <author>Europagate, 1995
10 <date>$Revision: 1.9 $
12 This document describes a Email server that provides access to the
21 This document describes an email server subsystem developed
22 within the EUROPAGATE project. The first part of this document
23 serves as an administrators guide, while the second part is a
24 follow-up on the Design deliverable (WP4.1) that outline the
25 deviations from the design. Also, the second part contains
26 a quick overview of the source code.
31 An ANSI C compiler is required in order to compile the ES software.
33 The ES can use either CNIDR's Zdist package or the YAZ package from
34 Index Data to interface the Z39.50 protocol. So you need to obtain
37 The zdist package can be found in:
40 url="ftp://ftp.cnidr.org/pub/NIDR.tools/zdist/zdist102b1-1.tar.Z"
41 name="ftp://ftp.cnidr.org/pub/NIDR.tools/zdist/zdist102b1-1.tar.Z">
43 The zdist package doesn't support result-set references. Also, it has a few
44 bugs. Therefore we've included a patch <tt/zdist.patch/ which fixes
46 Run patch in the directory above <tt/zdist102b1-1/:
51 The ES server only depends on <tt>libz3950.a</tt> so you only need
52 to build the zdist software in the directory <tt/libz3950/.
54 YAZ can be found at the FTP host:
56 <htmlurl url="ftp://ftp.indexdata.dk/index/yaz"
57 name="ftp://ftp.indexdata.dk/index/yaz">
59 The ES also use GNU's regex package to parse regular expressions.
60 The ES has been tested with regex-0.12. Some systems, such as Linux,
61 come with the regex package preinstalled.
63 Unpack <tt>egate.tar.gz</tt> and edit the top level <tt/Makefile/. Specify
64 where the GNU regex package is located and specify whether you use
65 YAZ or zdist. One some systems, you may have to set the <tt/NETLIB/ as
68 The shell variables <tt/CC/ and <tt/CFLAGS/ are used by the
69 <tt/Makefile/ so you may modify these before compiling.
71 Now, type <tt/make email/.
76 If the compilation succeeds, you should install the software in some
78 Edit the <tt/Makefile/ and set EMAILLIBDIR to the installation
79 directory. Since, the ES is executed by the mail system, and not by a
80 user, this directory shouldn't be globally executable.
82 When satisfied, type <tt/make install.email/.
84 Three executables are installed in EMAILLIBDIR:
86 <tag/eti/ The email transport interface. This program receives
87 incoming mail, identifies the user, and delivers the mail request
88 to the monitor or kernel (depending on configuration).
89 <tag/monitor/ The monitor
90 is optional component. The main objective
91 of the monitor is to limit the number of simultanous running kernel
93 <tag/kernel/ The kernel process is the core of the ES. It parses
94 the user's requests and interfaces the Z39.50 protocols.
97 The <tt/sendmail/ or a similar program delivers the mail to the
98 <tt/eti/ program. The <tt/sendmail/ program usually runs as user
99 <tt/mail/ or some other special user name. We strongly suggest that
100 you create a special user and group for the ES software. In this case
101 you should use <tt/chmod/ to and set the 'set user ID on execution'
102 bits on the executable files and give that user read/write/execute
103 permissions in EMAILLIBDIR.
105 The mail system needs to know about the ES. Pick some name that serves
106 as the ES user and edit <tt/aliases/ used by your mail system (usually
107 <tt>usr/lib/aliases</tt>). Now add the following line:
109 <tt>es:"|/usr/local/lib/es/eti </tt><em>options</em><tt>"</tt>
111 In this example the mail user name is <tt/es/ and the EMAILLIBDIR is
112 <tt>/usr/local/lib/es</tt>.
114 The ES system can operate with or without the monitor. When using
115 the monitor the number of simultanous running kernels can be
116 controlled. If the <tt>eti</tt> program is started with
117 two dashes (<tt>--</tt>) it will operate without the monitor and
118 the options specified after the two dashes are transferred to the
121 <sect1>Running with the monitor
124 The monitor must be running at all times in this mode. You should
125 start the monitor in one of your boot scripts (rc). For example this
126 might be put in a boot script:
129 (cd /usr/local/lib/es; ./monitor -d -l mon.log -- -d -l kernel.log &)
132 Here the monitor is started with the options <tt>-d -l mon.log</tt>
133 and the options after the two dashes are transferred to the
134 kernel. In this mode, the eti should contact the monitor (and not
135 the kernel), so the following might be put in the aliases file:
138 es:"|/usr/local/lib/es/eti -c /usr/local/lib/es"
141 The eti sets current directory to the path specified by option <tt>-c</tt>.
143 <sect1>Running without the monitor
146 In this mode you should never start the monitor.
147 The eti will contact the kernel directly. The following line could
148 be put in your aliases file:
151 es:"|/usr/local/lib/es/eti -c /usr/local/lib/es -- -d -l kernel.log"
157 The eti program accepts the following options:
159 <tag><tt>-l </tt>log</tag> The log file. If absent stderr is used.
160 <tag><tt>-d</tt></tag> Turns on debugging.
161 <tag><tt>-c </tt>dir</tag> Sets current directory to dir.
162 <tag><tt>-H</tt></tag> Help message.
163 <tag><tt>--</tt></tag> Indicates that the eti program should contact the
164 kernel (and not the monitor. All options after this one are transferred
171 The monitor program accepts the following command line options:
173 <tag><tt>-l </tt>log</tag> The log file. If absent stderr is used.
174 <tag><tt>-d</tt></tag> Turns on debugging.
175 <tag><tt>-H</tt></tag> Help message.
176 <tag><tt>--</tt></tag> Precedes options that are transferred to the kernel
179 The monitor normally reads the resource <tt>default.res</tt> in
180 current directory. You can change this behaviour by specifying an
181 alternate file on the command line.
186 List of options observed by the kernel:
188 <tag><tt>-d</tt></tag> Turns on debugging.
189 <tag><tt>-t </tt>target</tag> Opens connection to target (for testing only).
190 <tag><tt>-g </tt>lang</tag> Set language name.
191 <tag><tt>-o </tt>res</tag> Overriding resource file name. These
192 resources override both <tt>default.res</tt> and all user resources.
193 <tag><tt>-h </tt>host</tag> Override host name (for testing only).
194 <tag><tt>-p </tt>port</tag> Override port no (for testing only).
195 <tag><tt>-l </tt>log</tag> Specify log file.
196 <tag><tt>-H</tt></tag> Help message.
199 The kernel normally reads the resource <tt>default.res</tt> in
200 current directory. You can change this behaviour by specifying an
201 alternate file on the command line.
203 <sect>Managing the system
205 <sect1>Summary of files
208 To maintain the ES you need to know the files it uses. These are:
210 <tag>*.res</tag> Resource files with several settings that control
211 how the system operates, such as definition of targets, messages, etc.
212 <tag>*.bib</tag> Bib-1 attribute mapping files. These files describe
213 the mapping between CCL and the RPN query.
214 <tag>user.db</tag> Database of users. Only the eti process accesses
216 <tag>user.*.r</tag> Resource file for a user — accessed by the kernel
217 — only created when the user uses the <tt>def</tt> command.
218 <tag>user.*.p</tag> Persistency file for a user — accessed by
223 The ES system is mostly managed by resource files. The following
224 are example resource files that comes with the ES:
226 <tag><tt>default.res</tt></tag> General resource with reasonable defaults.
227 This file is read by the monitor and the kernel.
228 <tag><tt>loc.res</tt></tag> Resource file for Library of Congress test
230 <tag><tt>drewdb.res</tt></tag> Resource file for Data Research's test
232 <tag><tt>lang.uk.res</tt></tag> Resource file for english conversation.
233 <tag><tt>lang.dk.res</tt></tag> Resource file for danish conversation.
239 Most general resources should be set in the file <tt>default.res</tt>.
240 Some of the resources may be changed (overridden) by the user, while
241 others may be overridden by individual target defintions.
242 The complete scenario is depicted below:
248 |<---------| "target.res" |
252 |<---------| user.x.res |
256 |<---------| "lang.res" |
260 |<---------| "override" |
265 The following describes the general resources:
267 <tag>gw.reply.mta</tag> Name of MTA program — default
268 <tt>/usr/lib/sendmail</tt>.
269 <tag>gw.reply.tmp.prefix</tag> Prefix of temporary files used by the ES.
270 <tag>gw.reply.tmp.dir</tag> Name of directory with temporary files.
271 <tag>gw.marc.log</tag> If this resource is specified, retrieved MARC
272 records will be appended to this file.
273 <tag>gw.timeout</tag> Idle time before the kernel exits. When the
274 kernel exits, the Z39.50 persistency layer will reconnect when
276 <tag>gw.resultset</tag> If this setting is 1, the Z39.50 client will
277 use named result sets. If 0, the Z39.50 system will always use
278 <tt/Default/ as result-set name.
279 <tag>gw.persist</tag> If this setting 1, the persistency is enabled;
281 <tag>gw.max.process</tag> This settings is the maximum number of
282 simultaneous kernel processes — only used by the monitor.
283 <tag>gw.ignore.which</tag> Some targets doesn't indicate whether
284 a record is a diagnostic messaage or a database record. If this
285 setting is 1, the ES will always try to interpret the record as a
286 database record in ISO2709 format. If 0, the ES will use the
288 <tag>gw.default.show</tag> Default number of records to retrieve and display
289 when using the show command. This setting may be changed by the user
290 with the <tt>def defaultshow</tt> command.
291 <tag>gw.max.show</tag> This setting specifies the maximum number of
292 records the user may retrieve in one show command — default 100.
293 <tag>gw.autoshow</tag> Number of records to retrieve in a find
294 command — default 0. This setting may be changed by the user by
295 the <tt>def autoshow</tt> command.
296 <tag>gw.display.format</tag> Default display format. This setting may
297 be changed by the user by the <tt>def f</tt> command.
298 <tag>gw.language</tag> Current language. This setting may be
299 changed by the user with the <tt>def lang</tt> command. When the
300 langauge is set to something, say x, then the resource gw.lang.x
301 should hold a name of a resource file read by the kernel.
302 <tag>gw.lang.<em/x/</tag> Specifies name of resource file for
304 <tag>gw.target.<em/name/ </tag> Name of resource file of target
306 <tag>gw.portno</tag> Z39.50 target port number — default 210.
307 <tag>gw.hostname</tag> Z39.50 target host name.
308 <tag>gw.bibset</tag> Name of file with Bib-1 attribute mapping.
309 <tag>gw.databases</tag> Available databases on target.
310 <tag>gw.description</tag> Description of a target. This message
311 is returned to the user when the connection is established with the
313 <tag>gw.account</tag> Z39.50 Authentication string — default
320 There are several resource settings that deal with language
321 dependencies. These fall into the following categories that
322 depend on the resource name prefixes:
324 <tag>gw.msg</tag> Miscellaneous messages.
325 <tag>gw.err</tag> Error messages.
326 <tag>gw.bib1.diag.<em/no/</tag> Diagnostic error message indicated by
328 <tag>gw.help</tag> Help/description of various commands.
329 <tag>ccl.command</tag> CCL command names.
330 <tag>ccl.token</tag> CCL tokens names.
333 Refer to the sample files, <tt>default.res</tt>, <tt>lang.uk.res</tt>
334 and <tt>lang.dk.res</tt> for all available settings.
336 <sect1>Target definitions
339 To add a target definition called <em/mytarget/ you need to make a resource
340 entry in <tt>default.res</tt> called <tt>gw.target.</tt><em>mytarget</em>.
341 The value of this resource is the name of a resource file — for
342 example <em>mytarget</em><tt>.res</tt>. The resource file should at least
343 define the resources: <tt/gw.hostname/, <tt/gw.databases/ and
344 <tt/gw.description/. You might also consider specifying
345 <tt/gw.account/, <tt/gw.bibset/, <tt/gw.resultset/ and <tt/gw.portno/
346 in the target resource file. The user only needs to use the command
347 <tt>target </tt><em>mytarget</em> to use the target. Also, since we
348 already specified database names, the user doesn't need to use the
351 <sect1>CCL to RPN mapping
354 The mapping between CCL-queries and RPN are stored in files —
355 normally with the suffix <tt>.bib</tt>. We will refer these
356 files as bibset-files. You might consult the file <tt/default.bib/
357 to see an example of such file.
359 The mapping is necessary because targets usually only support a little
360 subset of the Bib-1 attribute set and because the CCL qualifiers
361 (field names) are not standardized. A bibset-file is specified
362 by the <tt/gw.bibset/ resource.
364 Column zero of a bib-file line either hold a hash character (<tt/#/)
365 indicating a comment in which case the rest of the line is
366 ignored; or a CCL qualifier.
368 The name of the CCL qualifier is up to you. However, the special
369 qualifier name <tt/term/ applies to the case where no qualifier
370 is specified in CCL. The CCL qualifier is
371 followed by one or more mapping specifications. A mapping
372 specification takes the form:
374 <em/type/<tt/=/<em/value/<tt/,/<em/value/...
376 The type is simply one of the six Bib-1 attribute query types:
378 <tag/u/ Use attribute. Value is an integer.
379 <tag/t/ Truncation attribute. Value is an integer; or the
380 value is a combination of:
382 <tag/l/ This character indicates that the CCL parser should allow
383 left truncation (2) if indicated by a <tt/?/ on the left side
385 <tag/r/ This character indicates that the CCL parser should allow
386 right truncation (1) if indicated by a <tt/?/ on the right side
388 <tag/b/ This character indicates that the CCL parser should allow
389 both left and right (3) truncation indicated by a <tt/?/ on both
390 left and right side of a term.
391 <tag/n/ This character indicates that the CCL parser should announce
392 no truncation (100) if no truncation was specified.
394 <tag/p/ Position attribute. Valus is an integer.
395 <tag/s/ Structure attribute. Value is an integer; or the
396 value is <tt/pw/ in which case the CCL parser announces word (2) or
397 phrase (1) depending on the number of adjacent terms.
398 <tag/r/ Relation attribute. Value is an integer; or the value is
399 <tt/o/ in which case, the CCL parser will select <em/less than/,
400 <em/less than or equal/, ... <em/greater than/ — depending on
401 the relation specified in CCL.
402 <tag/p/ Position attribute. Value is an integer.
405 Consider these bibset-lines:
411 The first line describes the mapping in when no qualifiers are
416 In this case the right truncation is enabled and the structure is
419 The second line is used in this search:
423 where the use attribute is <em/author/ and the structure is <em/word/.
425 The third line is used in:
429 where the use attribute is <em/date/ and the relation is <em/greater than/.
434 The implementation of the email server includes all the modules described
435 in the design deliverable.
437 The work was roughly carried out as follows:
439 <item>The logging facilities and resource management utilities were
440 implemented — virtually all other modules depend on these
442 <item>A minimal ES was implemented — including a high-level
443 API to the Z39.50 sub-system and a CCL parser with a few
444 commands, such as FIND and SHOW. This version displayed MARC
445 records in a raw format. This version served as base for the URP.
446 <item>The first version of the MARC display formatting tool, FML,
447 was implemented and included in the ES.
448 <item>The ETI program was implemented along with the IPC
449 (interprocess communication) utilities based on FIFOs. Facilities
450 to keep connections alive (to Z39.50 targets) was implemented.
451 To identify a user, a file-resident symbol table (small database) was
452 implemented which maps a email username to a unique integer (email userid).
453 <item>The protocol persistency was implemented and more CCL commands
455 <item>The monitor program was implemented.
458 The following sections cover the most important modules in the ES and
459 deviations from the design.
461 <sect1>Z39.50 Interface layer
464 The design report specified that the Zdist toolkit from CNIDR would
465 be used in the ES to provide access to the Z39.50 protocol. The package
466 was choosen bacause it is easy to use and, more important, we felt
467 that the API would be reasonably stable and supported.
469 Nevertheless it turned out that CNIDR choose to change the API
470 completely around January 1995 and announced a new version
473 <em>Note: As of this date the newest version of Zdist is still
474 zdist102b1-1. CNIDR seems to concentracte on their Isite package
475 which also includes a Zdist package presumably similar to the
476 standalone Zdist package</em>
478 During the work with the Zdist package a few bugs were discovered.
479 Fortunately, they could be solved within a few days. We also
480 discovered that the package lacks result-set references.
481 We posted the bug fixes to Kevin Gamiel who is responsible for
482 the package but we didn't get responses. So, eventually, we weren't
483 satisfied with the package after all.
485 In February some of us began the development of a new Z39.50 package
486 called YAZ — in retrospect somewhat motivated by the
487 experiences with existing Z39.50/SR toolkits.
489 To support result-set references we chose to incorporate a YAZ
490 interface in the ES also. And we designed and implemented a
491 simple high-level Z39.50 origin API that supported both Zdist and YAZ.
493 The protocol persistency module was implemented on top of
494 the high-level API and not on top of Zdist. The obvious
495 advantage is that the persistency module is not tied to one
496 particular Z39.50/SR package.
498 Persistency information stored for each user is simply:
500 <item>hostname and port number.
501 <item>authentication string
502 <item>selected database(s)
503 <item>next result set number
504 <item>next result set position
505 <item>result set information
508 Information about each result set includes:
511 <item>size (number of hits)
516 A persistency file is removed each time a new target is selected.
517 It is our experiences that the persistency files are very small.
522 The CCL was implemented as described in the design. A CCL utility
523 was made as a separate module which implements a tokenization
524 package and a parser which translates from FIND to RPN. The
525 data structure used to represent the RPN query is also used in
526 Z39.50 search API on top of YAZ or Zdist.
528 The CCL parser is quite configurable. Token names can be redefined to
529 one or more names (aliases). Also, the specification of mapping
530 between CCL field names (qualifiers) and Bib-1 attributes can be
531 specified in either the C API or a file.
533 Although the Z39.50 system in the ES uses the Bib-1 attribute set, the
534 CCL parser itself is not tied to Bib-1.
539 The FML system is used to handle the presentation of MARC
540 records. There are some deviations to the design report, however.
541 The most important changes are:
543 <item>The <tt/expr/ function is not implemented. Instead arithmetic
544 operators <tt/plus/, <tt/minus/, <tt/mult/ and <tt/div/ are
545 implemented. Also relational operators <tt/gt/, <tt/lt/ ... are
547 <item>The <tt/lindex/ function is called <tt/index/ and it is a binary
548 operator where the left operand is the list and the right operand is
550 <item>The MARC extraction routines are not implemented.
551 Instead, a MARC record is transferred as an argument
552 to a formatting-routine (in list notation). The formatting
553 routine then extracts fields from the list by list/string
554 manipulation functions.
555 <item>A new statement, <tt/bin/, is implemented to define
556 binary operators (functions).
562 As described in the design, FIFOs are used to communicate between
563 the ETI, monitor and kernel. The ES can run without the monitor,
564 however. The primary reason for the presence of the monitor was
565 to assure that the kernel releases the resources used by the
566 persistency layer. But, since the persistency layer did turn out to
567 use virtually no disk space at all, there was no point in starting
568 a kernel process to remove its files — hence this facility
569 was not implemented. The only purpose of the monitor is to keep the
570 number of running kernels at a maximum level and even that
571 is probably useless since most unices will swap kernel processes
575 before a kernel exits and saves its persistency file is not
576 controlled by the monitor. Saving the persistency file and
577 keeping it is usually a good approach — even when a
578 user doesn't reference/show old result-sets since the user
579 has a notion of <em/current target/ and database.
584 In this section a short description of each source module is
585 given. Each module is implemented in a separate sub directory.
586 Any public headers are located in the <tt/include/ directory.
589 <tag/res+log/ is an implementation of the logging system
590 and the resource management sub system. Note that the
591 resource module depends on the logging facility. Logging
592 is implemented in <tt>gw-log.c</tt> and <tt/gw-log.h/. The
593 file <tt>gw-log-test.c</tt> is small test program for the
594 logging system. The core of the resource management is implemented
595 in <tt>gw-res.c</tt>. The files <tt>gw-res-bool.c</tt> and
596 <tt>gw-res-int.c</tt> implement two utility routines &mdash
597 on top of the resource management. The header file
598 <tt>gw-resp.h</tt> is a private header file and <tt>gw-res.h</tt>
599 is a public header file.
601 <tag/ccl/ implements CCL to RPN mapping and a tokenization
602 utility for other CCL commands. The mapping function is
603 implemented in <tt>cclfind.c</tt>. Qualifiers are handled in
604 <tt>cclqual.c</tt> while reading of qualifier mappings from a
605 file is implemented in <tt>cclqfile.c</tt>. Scanning is implemented
606 in <tt>ccltoken.c</tt>. String utilities, which might be changed if
607 other character sets are needed, is implemented in
608 <tt>cclstr.c</tt>. Table of error messages is implemented in
611 <tag/util/ implements various utilities:
613 <tag>MARC utility</tag> implemented in <tt>iso2709</tt>...
614 <tag>Database utility</tag> implemented in <tt>gw-db.[ch]</tt>. This
615 utility is used to map a user (email) to an integer.
616 <tag>String queue utility</tag> implemented in <tt>strqueue.[ch]</tt>. This
617 utiltiy is used to queue incoming mail in the ETI, kernel and
619 <tag>Pretty printer</tag> implemented in <tt>ttyemit.[ch]</tt>
620 — used by the URP.
621 <tag>FIFO IPC utiltiy</tag> implemented in <tt>gip*.[ch]</tt> —
622 used by the ETI, kernel and monitor.
625 <tag/fml/ implements FML. The top level functions are implemented
626 in <tt>fml.c</tt>, <tt>fmlcall.c</tt> and <tt>fmlcalls.c</tt>.
627 Scanning is implemented in <tt>fmltoken.c</tt>.
628 Memory management is implemented in <tt>fmlmem.c</tt>.
629 Arithmetic operators are implemented in <tt>fmlarit.c</tt>.
630 String manipulation functions are implemented in <tt>fmlstr.c</tt>.
631 Relational operators are implemented in <tt>fmlrel.c</tt>.
632 List maniuplations are performed in <tt>fmllist.c</tt>.
633 FML symbol table management is implemented in <tt>fmlsym.c</tt>.
634 Conversion from ISO2709 to list notation is implemented in
637 <tag/zlayer-zdist/ implements the high-level Z39.50 API on top
638 of Zdist. This task is implemented in <tt>zaccess.c</tt>. The
639 public header file is called <tt>zaccess.h</tt>.
641 <tag/zlayer-yaz/ implements the high-level Z39.50 API on top
642 of YAZ. This task is implemented in <tt>zaccess.c</tt>. The
643 public header file is called <tt>zaccess.h</tt>.
645 <tag/kernel/ implements the ETI, kernel and monitor. The kernel
646 itself is implemented in <tt>main.c</tt>, <tt>urp.c</tt> and
647 <tt>persist.c</tt>. The ETI is implemented in <tt>eti.c</tt> and
648 the monitor is implemented <tt>monitor.c</tt>.
654 Copyright © 1995, the EUROPAGATE consortium (see below).
656 The EUROPAGATE consortium members are:
659 <item>University College Dublin
660 <item>Danmarks Teknologiske Videnscenter
661 <item>An Chomhairle Leabharlanna
662 <item>Consejo Superior de Investigaciones Cientificas
665 Permission to use, copy, modify, distribute, and sell this software and
666 its documentation, in whole or in part, for any purpose, is hereby granted,
669 1. This copyright and permission notice appear in all copies of the
670 software and its documentation. Notices of copyright or attribution
671 which appear at the beginning of any file must remain unchanged.
673 2. The names of EUROPAGATE or the project partners may not be used to
674 endorse or promote products derived from this software without specific
675 prior written permission.
677 3. Users of this software (implementors and gateway operators) agree to
678 inform the EUROPAGATE consortium of their use of the software. This
679 information will be used to evaluate the EUROPAGATE project and the
680 software, and to plan further developments. The consortium may use
681 the information in later publications.
683 4. Users of this software agree to make their best efforts, when
684 documenting their use of the software, to acknowledge the EUROPAGATE
685 consortium, and the role played by the software in their work.
687 THIS SOFTWARE IS PROVIDED "AS IS" AND WITHOUT WARRANTY OF ANY KIND,
688 EXPRESS, IMPLIED, OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY
689 WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
690 IN NO EVENT SHALL THE EUROPAGATE CONSORTIUM OR ITS MEMBERS BE LIABLE
691 FOR ANY SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF
692 ANY KIND, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA
693 OR PROFITS, WHETHER OR NOT ADVISED OF THE POSSIBILITY OF DAMAGE, AND
694 ON ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
695 USE OR PERFORMANCE OF THIS SOFTWARE.