The WN server has powerful capabilities for modifying a text
file as it is served. One such feature, called "server-side includes",
is the ability to automatically insert one file in another. But the
usefulness of this capability is greatly enhanced by another feature, conditional text. A simple "if - else -
endif
" construct allows sections of a document to be served only
to certain hosts, or to clients providing an appropriate
Accept
header or a desired Referer
or
User-Agent
header. Using the "else
" construct
allows alternate text segments for clients not meeting the desired
criteria. For these features to work it is necessary for the server to
parse the file looking for the location of includes or for conditionally
served text. The fact that a file should be parsed in this way is
indicated in its entry in its index
file. This is done with an
entry "Attributes=parse
"
in the file record. This line is not necessary if the file record lists
wrappers or includes
since it is assumed in that case.
The simplest and most common use of parsed text is including additional
files in one which is being served or "wrapping" a served file with a
second file, i.e. inserting the requested file inside the second at a
desired location. The latter is useful, for example, if you wish to
place a standard message at the beginning or end (or both) of a large
collection of files. All files included or used as wrappers must be
listed in the index.cache
file. They are not listed in the
file in which they are to be included; only the location of the insertion
is marked.
To ensure security various options are available including the the
requirement that a served file and all its includes and wrappers have the
same owner as the index.cache
file listing them. This is
done with the -u
option.
Another important application of wrappers is to customize the HTML documents returned listing the successful search matches. If, for example, a directory is assigned a wrapper the server assumes that it contains all text describing the search and it merely supplies an unordered list of links to the matching items.
Suppose your server offers the file foo.html
and you wish to
have the file bar.html
automatically inserted at some point
into foo.html
. You achieve this by use of the "Includes=
" directive in
the index
file in the
directory containing foo.html
. That index
file should have an entry
like:
File=foo.html
Includes=bar.html
and the file foo.html
should contain either of the lines:
<!-- #include -->
or:
<?WN #include >
These two forms are equivalent as far as the server is concerned. The
second form is considered more SGML friendly by many as
<?WN something>
indicates a processing
instruction specific to WN rather than a comment. For
historical reasons this manual describes the other form, but either may
be used. With the first form the '#
' is required but with
the other you may use either:
<?WN #include>
or:
<?WN include>
Any of the "<!-- #something -->
" lines
described in this section may be replaced using the
"<?WN something>
" syntax. Also with this syntax
the case of the WN is not significant.
The marker "<!-- #include -->
" should be the
only thing on its line but may have white space before it. This is
controlled by the "Includes=
" line in the
index
file. All
including of files by WN is done only for text files and only in
units of lines.
If the include marker is never found in foo.html
, then this
file is served and the file bar.html
, is appended at the
end. Thus the "Includes=
" directive can
be used to append a file without the need of any marker in the main file.
If you wanted to include two files in foo.html
, say
bar.html
and bah.html
you would place the
marker "<!-- #include -->
" at two places in
foo.html
and have an entry in your index
file like:
File=foo.html
Includes=bar.html,bah.html
The two files will then be included at the marked spots in the order that
they are listed in the "Includes=
" directive,
i.e. bar.html
will be inserted at the first marker and
bah.html at the second. You can have as many included files as you wish.
They should all be listed on the "Includes=
" line separated
by commas.
Sometimes it is easier to specify the order in which multiple files are
to be inserted by adding the file name in the HTML document. This is also
possible with WN. However, since good security requires that
only files listed in the index
file may be served, it is
still necessary to list files you want to include there.
So, another way to include two files in foo.html
, say,
bar
and bah
, is to place the markers
'<!-- #include "bar" -->
' and
'<!-- #include "bah" -->
' at the
appropriate places in foo.html
and have an entry in your
index
file like:
File=foo.html
List-Includes=bar,bah
The order of the files listed in the directive is not significant. The
example above grants permission for the inclusion of the three files
listed. It does not require their insertion. Note the difference
between the two mechanisms: with the "Includes=
" directive only
"<!-- #include -->
" is needed in the HTML and the file inserted is taken
(in order) from the list in the directive, while with the "List-Includes=
"
directive the HTML must contain
'<!-- #include "bah" -->
' so the server
knows to insert file "bah
" and the file list in the
directive is checked to see that permission to serve this file has been
granted. If you use the "List-Includes=
"
directive you cannot also use either the "Includes=
" or "Wrappers=
" directives --
the "List-Includes=
" and
"Includes=
"/"Wrappers=
" mechanisms are
mutually exclusive. It is fine to use the line
'<!-- #include "bar.html" -->
' to
insert a file with the "Includes=
" directive, but
with that directive the "bar.html
" is only a comment and has
no effect on which file is actually inserted.
There is a "Default-List-Includes=
"
directory directive which can be used to grant permission for any HTML file in the directory to
include one or more of the listed files. This also causes all HTML files in the directory to
acquire the "Attributes=parse
".
If a file has a filter only that file will be filtered, not any wrappers or includes.
Instead of a file it is possible to include the output of a program. To
do this the program is listed in the "Includes=
" directive but
its name is preceded with the '!
' character. For example:
File=foo.html
Includes=!/usr/bin/date,bar.html
will insert the time and date at the first
"<!-- #include -->
" and the contents of
bar.html
at the second. "Includes=
" and "List-includes=
"
should not list arguments to programs. Thus if you wished to use the
output of the UNIX date
utility to get UTC instead of local time you should create a program
named "mydate
" containing, for example:
#!/bin/sh
exec /usr/bin/date -u
and use the directive "Includes=mydate
".
This ability to serve the output of programs can be restricted in several
ways. If WN is invoked with the -e
option then no includes,
filters, or CGI
programs will be executed.
The -E
option in
conjunction with the -t
or -T
options restricts
the execution of programs to those listed in index.cache
files owned by trusted users or groups. The -u
option allows the
execution of programs or inclusion of files owned by the owner of the
index.cache
file which lists them. If the -E
and -u
options are used
together the -E
option
takes precedence.
If you wish to have all the standard CGI/1.1 environment variables made available to the executed include program you can do so by adding the line:
Attributes=cgi
to the file record. A list of these environment variables can be found in the appendix "CGI and other Environment Variables for the WN Server" in this guide. Also see the sample CGI program which is located in the file /docs/examples/sample.cgi which accompanies the WN distribution.
Often a server maintainer may wish to serve different versions of a document to different clients. Here is a simple example of how this can be handled with WN.
<!-- #if accept =~ "image/jpeg" -->
<a href="picture.jpg">
Here is the jpeg version of the picture:
</a>
<!-- #else -->
<a href="picture.gif">
Here is the gif version of the picture:
</a>
<!-- #endif -->
This tells the server to look at the Accept: headers provided by the
client and if "image/jpeg
" is among them then use the
jpeg
image and otherwise use the gif
image.
More precisely, the "image/jpeg
" part of the
"if
" line is a UNIX grep(1)
utility like regular expression and if there is any match for it among
the Accept
headers the jpeg
image will be used.
Of course these "if - else - endif
" constructs can be
nested. A similar construct allows you to make the text served depend on
matching a regular expression with the contents of the
User-Agent
header, the Referer
header, the
Cookie
header, the client hostname, or the client IP
address. A complete list of possible test clauses for the
#if
statement is contained in appendix "Parsed Document Syntax for the WN
Server" in this guide.
Note: For this to work the file containing the conditional text needs to be parsed by the server. The server only knows to do this if the file record in theindex
file contains a line "Attributes=parse
".
If, in an "if clause", instead of the equal-tilde string
'=~
' (to indicate a regular expression match) the character
string '!~
' is used then the truth value of the match is
reversed. For example the lines:
<!-- #if referer !~ "my.host.edu" -->
Here is some text for newcomers to my site.
<!-- #endif -->
would display the "Here is some text ...
" to those clients
accessing this document via any link which is not on the host
my.host.edu
.
There is also a redirect command which can conditionally cause the server to send an HTTP/1.1 redirect to a new URL. For example if the text:
<!-- #if hostname =~ "\.uk$" -->
<!-- #redirect = "UK_mirror_url" -->
<!-- #endif -->
is included at the beginning of an HTML document then any request from
a UK host will automatically be redirected to the specified URL, the
UK_mirror_url
in this case. This mechanism could also be
used to redirect text only browsers to a text only alternative page, etc.
There must be no text sent before the
'<!-- #redirect = "url" -->
' is
encountered (not even blank lines) since the server cannot send an HTTP/1.1 redirect while in the
middle of transmitting a document.
Normally the URL in the
'<!-- #redirect = "URL" -->
' line
is fully qualified, like "http://host/path/foo
". However,
it can also be simply "foo
" referring to a file in the same
directory as the file being parsed. In this case an HTTP/1.1 redirection is not
sent, and instead the file "foo
" is returned immediately to
the client.
Often a single regular expression is not adequate to distinguish whether
or not to serve some text. For that reason WN allows you to use
a file containing any number of regular expressions and serve a document
based on whether any of these expressions match accept headers, referer
header, user-agent, hostname, etc. For example, if the file
"acceptfile
" contains the lines:
image/gif
image/jpeg
image/x-xbitmap
then the following conditional text might be appropriate:
<!-- #if accept file = "acceptfile" -->
I see you aren't using a text only browser...
<!-- #endif -->
The format of the file "acceptfile
" is one UNIX grep(1)
utility like regular expression per line. Lines beginning with
'#
' are taken to be comments. If a regular expression is
preceded by the character '!
' then that character is skipped
but the truth value of the match is reversed. More information about
files of regular expressions for conditional text can be found in
appendix "Parsed Document Syntax for the
WN Server" in this guide.
Another method of using conditional text is with a normal WN access control file. For example:
<!-- #if accessfile="secret/access" -->
Here is a link to a restricted directory.
<a href="/secret/stuff.html">Restricted Local Stuff</a>
Hosts not listed in the file "secret/access" can't look at it, so why
show them a link to it?
<!-- #endif -->
will display the included text and the link to
"Restricted Local Stuff
" only to clients on hosts
permitted by the access control file
"secret/access
".
The condition in the "if
" or "elif
" tags can be
made more complex than those described above by combining simple
conditions using the logical operations '&&
' for
'and', '||
' for 'or' and '!
' for 'not'.
Parentheses may be used for grouping. For example:
<!-- #if hostname =~ "\.uk$" && accessfile="secret/access" -->
Something for those sites in accessfile but not in the UK.
<!-- #endif -->
This is described in greater detail in the section "Matching Regular Expressions:
#if
and #elif
" in this guide.
section
MarkerOften you may want to include HTML files in a larger document. For example, a single logical HTML document often consists of a number of separate files. This can make it very difficult for a client to download and print the entire document. To alleviate this problem you could form a single document concatenating all the pieces, but that can be problematic when the document is altered or updated as it is necessary to change both the piece and the concatenation.
WN offers a solution to this by allowing you to write a small
skeleton document which includes all the pieces, forming a "virtual"
document which is the concatenation. But if we just used the mechanism
above the concatenation document would include the
<head>
elements of all the pieces. To remedy this you
use the marker:
<!-- #section -->
instead of "<!-- #include -->
" and the
server will include only the portion of the HTML document between the special
comments "<!-- #start -->
" and
"<!-- #end -->
" inserted in that document.
This requires that these starting and ending comments occur in the HTML document on lines by
themselves.
Suppose you have a large number of files and want a standard header or
footer placed on all of them (perhaps a standard disclaimer). You could,
of course, make many copies of the standard header with a different
"Includes=
"
directive for each, but this is cumbersome. It has other drawbacks as
well, such as the fact that grep and context searches do not search associated
include files, only the main file.
To deal with this case WN uses wrappers. Wrapping a
file is the inverse of including it. If you have an index
file entry like:
File=foo.html
Wrappers=bar.html
then the server will send the file bar.html
looking for the
marker "<!-- #include -->
" and inserting
foo.html
at the line where it is found. So this is just
like the "Includes=
" directive
except the role of which file is included in the other is reversed. If
the include marker is never found the entire wrapper,
bar.html
, is sent first and the main file,
foo.html
, is appended at the end. Thus the "Wrappers=
" directive can
be used to prepend a file without the need of any marker in the main
file.
A search wrapper is a special kind of wrapper. Its function is to allow you to customize the results of user searches. It is normally associated with an entire directory in which case it applies to all title, keyword, field, grep and context searches of that directory. See the chapter "Setting Up Searches on the WN Server" in this guide.
A line like:
SearchWrapper=foo.html
in the directory record of an index
file will cause any search
of that directory to return an unordered list of matches wrapped with the
file foo.html
. The list of matches will be inserted into
foo.html
at a point where the marker
"<!-- #include -->
" is found. You can also
insert the user supplied search term by using the marker
"<!-- #query -->
". Both of these markers
must occur on a line by themselves.
Wrapping and including can be arbitrarily nested (though this is not true of search wrappers). It can be a little complicated to get the effect you want.
Here is how to do it. Think of all your files, wrappers, includes and
the main file arranged as you wish them to be combined for the final
served document. Now imagine inserting an opening (or left) parenthesis
at the beginning of each file and insert a closing (or right) parenthesis
at the end of each file. You should have a legally nested and balanced
collection of parentheses. To each of the opening parentheses attach the
name of the file which begins at that point. Then write down the list of
all the file names in the order their corresponding opening parentheses
occur. All the files which come before the main file should be wrappers
and should be listed in the "Wrappers=
" line in the
order in which they occur in this list. All the files after the main
file should be in the "Includes=
" line and should
occur in the order they occur in this list.
Here's a simple example. Suppose we have a main file M
and
other files A
, B
, C
, and
D
which we want to have nested like:
(D...(B...B)...(M...(A...(C...C)...A)...M)...D)
Then the entry in the index
file should look like:
File=M
Wrappers=D,B
Includes=A,C
Often it is useful to insert items like the main document's title into a wrapper, or the search item provided by the client into a searchwrapper for the responses. This is possible with WN parsed documents using the lines:
<!-- #title -->
or:
<!-- #query -->
in the file at the point where the title or search term should be inserted. Both of these markers must occur on a line by themselves.
In addition to the title and query user supplied fields can be inserted in documents. This is done by including a marker such as:
<!-- #field3 -->
in the file at the point where the value should be inserted.
You can also insert the value of any environment variable into your text with a line like:
<!-- #environ = "HTTP_REFERER" -->
which will be replaced by the contents of the environment variable
HTTP_REFERER
.
Note: If there is no "Wrappers=
" or "Includes=
" line in theindex
file for this entry then there must be a line like:
Attributes=parse
so the server knows it is to parse the file to look for the marker. See
the section "Set File
Attributes" in this guide for more details. Also if you wish to
insert the value of a CGI/1.1 environment variable
(as in the example above) you must have an "Attributes=parse
"
line.