Sophie

Sophie

distrib > Mageia > 4 > x86_64 > by-pkgid > e7cba66dcc6c6829ee4a34287aca1aae > files > 10

lib64xpa-devel-2.1.15-3.mga4.x86_64.rpm

<!-- =defdoc xpaclient xpaclient 3 -->
<HTML>
<HEAD>
<TITLE>XPA Client API</TITLE>
</HEAD>
<BODY>

<!-- =section xpaclient NAME -->
<H2><A NAME="xpaclient">XPAClient: The XPA Client-side Programming Interface</A></H2>

<!-- =section xpaclient SYNOPSIS -->
<H2>Summary</H2>
A description of the XPA client-side programming interface.

<!-- =section xpaclient DESCRIPTION -->
<H2><A NAME="intro">Introduction to XPA Client Programming</H2></A>
<P>
Sending/receiving data to/from an XPA access point is easy: you
generally only need to call the XPAGet() or XPASet() subroutines.
<PRE>
  #include &lt;xpa.h&gt;

  int <A HREF="./client.html#xpaget">XPAGet</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      char **bufs, size_t *lens, char **names, char **messages, int n);

  int <A HREF="./client.html#xpaset">XPASet</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      char *buf, size_t len, char **names, char **messages, int n);

  int <A HREF="./client.html#xpainfo">XPAInfo</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      char **names, char **messages, int n);

  int <A HREF="./client.html#xpaaccess">XPAAccess</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      char **names, char **messages, int n);

  int <A HREF="./client.html#xpagetfd">XPAGetFd</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      int *fds, char **names, char **messages, int n);

  int <A HREF="./client.html#xpasetfd">XPASetFd</A>(XPA xpa,
      char *template, char *paramlist, char *mode,
      int fd, char **names, char **messages, int n);

  XPA <A HREF="./client.html#xpaopen">XPAOpen</A>(char *mode);

  void <A HREF="./client.html#xpaclose">XPAClose</A>(XPA xpa);

  int <A HREF="./client.html#xpanslookup">XPANSLookup</A>(XPA xpa,
      char *template, char *type,
      char ***classes, char ***names, char ***methods, char ***infos);
</PRE>

<H2>Introduction</H2>

To use the XPA application programming interface, a software developer
generally will include the xpa.h definitions file:
<PRE>
  #include &lt;xpa.h&gt;
</PRE>
in the software module that defines or accesses an XPA access point and
then will link against the libxpa.a library:
<PRE>
  gcc -o foo foo.c libxpa.a
</PRE>
XPA has been compiled using both C and C++ compilers.
<P>
Client communication with XPA public access points generally is
accomplished using XPAGet() or XPASet() within a program (or xpaget
and xpaset at the command line).  Both routines require specification
of the name of the access point.  If a <A HREF="./template.html">template</A>
is used to specify the access point name (e.g., "ds9*"), then
communication will take place with all servers matching that template.

<!-- =defdoc xpaget xpaget 3 -->

<!-- =section xpaget NAME -->
<H2><A NAME="xpaget">XPAGet: retrieve data from one or more XPA servers</A></H2>

<!-- =section xpaget SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPAGet(XPA xpa,
             char *template, char *paramlist, char *mode,
             char **bufs, size_t *lens, char **names, char **messages,
	     int n);
</PRE>
</B>

<!-- =section xpaget DESCRIPTION -->
<P>
Retrieve data from one or more XPA servers whose class:name identifier
matches the specified template.

<P>
A 
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most n matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, then the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPAGet() routine then retrieves data from at most n XPA servers,
places these data into n allocated buffers and places the buffer
pointers in the bufs array. The length of each buffer is stored in the
lens array. A string containing the class:name and ip:port is stored
in the name array.  If a given server returned an error or the server
callback sends a message back to the client, then the message will be
stored in the associated element of the messages array.  NB: if
specified, the name and messages arrays must be of size n or greater.

<p>
The returned message string will be of the form:
<PRE>
  XPA$ERROR error-message (class:name ip:port)
</PRE>
or
<PRE>
  XPA$MESSAGE message (class:name ip:port)
</PRE>
<P>
Note that when there is an error stored in an messages entry, the
corresponding bufs and lens entry may or may not be NULL and 0
(respectively), depending on the particularities of the server.

<P>
The return value will contain the actual number of servers that were
processed.  This value thus will hold the number of valid entries in
the bufs, lens, names, and messages arrays, and can be used to loop
through these arrays.  In names and/or messages is NULL, no information is
passed back in that array.

<P>
The bufs, names, and messages arrays should be freed upon completion (if
they are not NULL);

<P>
The mode string is of the form: "key1=value1,key2=value2,..."
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server (after callback completes)
  doxpa		true/false	true		client processes xpa requests
</PRE>
<P>
The ack keyword is not very useful, since the server completes the callback
in order to return the data anyway.  It is here for completion (and perhaps
for future usefulness).

<p>
Normally, an XPA client will process incoming XPA server requests
while awaiting the completion of the client request.  Setting this
variable to "false" will prevent XPA server requests from being
processed by the client.
					   
<P>
<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;

  #define NXPA 10
  int  i, got;
  size_t  lens[NXPA];
  char *bufs[NXPA];
  char *names[NXPA];
  char *messages[NXPA];
  got = XPAGet(NULL, "ds9", "file", NULL, bufs, lens, names, messages,
  NXPA);
  for(i=0; i&lt;got; i++){
    if( messages[i] == NULL ){
      /* process buf contents */
      ProcessImage(bufs[i], ...);
      free(bufs[i]);
    }
    else{
      /* error processing */
      fprintf(stderr, "ERROR: %s (%s)\n", messages[i], names[i]);
    }
    if( names[i] )
      free(names[i]);
    if( messages[i] )
      free(messages[i]);
  }
</PRE>

<!-- =defdoc xpaset xpaset 3 -->

<!-- =section xpaset NAME -->
<H2><A NAME="xpaset">XPASet: send data to one or more XPA servers</A></H2>

<!-- =section xpaset SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPASet(XPA xpa,
             char *template, char *paramlist, char *mode,
             char *buf, size_t len, char **names, char **messages,
             int n);
</PRE>
</B>

<!-- =section xpaset DESCRIPTION -->
<P>
Send data to one or more XPA servers whose class:name identifier
matches the specified template.

<P>
A 
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most n matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPASet() routine transfers data from buf to the XPA servers.
The length of buf (in bytes) should be placed in the len variable.

<P>
A string containing the class:name and ip:port of each of these server
is returned in the name array.  If a given server returned an error or
the server callback sends a message back to the client, then the
message will be stored in the associated element of the messages
array. NB: if specified, the name and messages arrays must be of size
n or greater.

<p>
The returned message string will be of the form:

<PRE>
  XPA$ERROR   [error] (class:name ip:port)
</PRE>
or
<PRE>
  XPA$MESSAGE [message] (class:name ip:port)
</PRE>
<P>
The return value will contain the actual number of servers that were
processed.  This value thus will hold the number of valid entries in
the names and messages arrays, and can be used to loop through these
arrays.  In names and/or messages is NULL, no information is passed back
in that particular array.

<P>
The mode string is of the form: "key1=value1,key2=value2,..."
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server (after callback completes)
  verify	true/false	false		send buf from XPASet[Fd] to stdout
  doxpa		true/false	true		client processes xpa requests
</PRE>
<P>
The ack keyword is useful in cases where one does not want to wait for
the server to complete, e.g. if a lot of processing needs to be done
by the server on the passed data or when the success of the server
operation is not relevant to the client.

<p>
Normally, an XPA client will process incoming XPA server requests
while awaiting the completion of the client request.  Setting this
variable to "false" will prevent XPA server requests from being
processed by the client.
					   
<P>
<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;

  #define NXPA 10
  int  i, got;
  size_t  len;
  char *buf;
  char *names[NXPA];
  char *messages[NXPA];
  ...
  [fill buf with data and set len to the length, in bytes, of the data]
  ...
  /* send data to all access points */
  got = XPASet(NULL, "ds9", "fits", NULL, buf, len, names, messages, NXPA);
  /* error processing */
  for(i=0; i&lt;got; i++){
    if( messages[i] ){
      fprintf(stderr, "ERROR: %s (%s)\n", messages[i], names[i]);
    }
    if( names[i] )    free(names[i]);
    if( messages[i] ) free(messages[i]);
  }
</PRE>
					   
<!-- =defdoc xpainfo xpainfo 3 -->

<!-- =section xpainfo NAME -->
<H2><A NAME="xpainfo">XPAInfo: send short message to one or more XPA servers</A></H2>

<!-- =section xpainfo SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPAInfo(XPA xpa,
              char *template, char *paramlist, char *mode,
	      char **names, char **messages, int n);
</PRE>
</B>

<!-- =section xpainfo DESCRIPTION -->
<P>
Send a short paramlist message to one or more XPA servers whose
class:name identifier matches the specified
<A HREF="./template.html">template</A>.

<P>
A
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most n matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, then the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPAInfo() routine does not send data from a buf to the XPA
servers. Only the paramlist is sent.  The semantics of the paramlist
is not formalized, but at a minimum is should tell the server how to
get more information.  For example, it might contain the class:name
of the XPA access point from which the server (acting as a client)
can obtain more info using XPAGet.

<P>
A string containing the class:name and ip:port of each server is
returned in the name array.  If a given server returned an error or
the server callback sends a message back to the client, then the
message will be stored in the associated element of the messages
array.  The returned message string will be of the form:
<PRE>
  XPA$ERROR   error-message (class:name ip:port)
</PRE>
or
<PRE>
  XPA$MESSAGE message 	  (class:name ip:port)
</PRE>
<P>
The return value will contain the actual number of servers that were
processed.  This value thus will hold the number of valid entries in
the names and messages arrays, and can be used to loop through these
arrays.  In names and/or messages is NULL, no information is passed back
in that array.

<P>
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server
</PRE>
<P>
When ack is false, XPAInfo() will not wait for an error return from the XPA
server. This means, in effect, that XPAInfo will send its paramlist string
to the XPA server and then exit: no information will be sent from the server
to the client. This UDP-like behavior is essential to avoid race
conditions in cases where XPA servers are sending info messages to
other servers. If two servers try to send each other an info message
at the same time and then wait for an ack, a race condition will result and
one or both will time out.

<P>
<B>Example:</B>
<PRE>
  (void)XPAInfo(NULL, "IMAGE", "ds9 image", NULL, NULL, NULL, 0);
</PRE>
					   
<!-- =defdoc xpagetfd xpagetfd 3 -->

<!-- =section xpagetfd NAME -->
<H2><A NAME="xpagetfd">XPAGetFd: retrieve data from one or more XPA servers and write to files</A></H2>

<!-- =section xpagetfd SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPAGetFd(XPA xpa,
               char *template, char *paramlist, char *mode,
	       int *fds, char **names, char **messages, int n);
</PRE>
</B>

<!-- =section xpagetfd DESCRIPTION -->
<P>
Retrieve data from one or more XPA servers whose class:name identifier
matches the specified
<A HREF="./template.html">template</A>
and write it to files associated with
one or more standard I/O fds (i.e, handles returned by open()).

<P>
A 
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most ABS(n) matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, then the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPAGetFd() routine then retrieves data from the XPA servers,
and write these data to the fds associated with one or more fds
(i.e., results from open). Is n is positive, then there will be n fds
and the data from each server will be sent to a separate fd. If n is
negative, then there is only 1 fd and all data is sent to this single
fd. (The latter is how xpaget is implemented.)

<P>
A string containing the class:name and ip:port is stored in the name
array.  If a given server returned an error or the server callback
sends a message back to the client, then the message will be stored in
the associated element of the messages array.  NB: if specified, the
name and messages arrays must be of size n or greater.

<P>
The returned message string will be of the form:
<PRE>
  XPA$ERROR   error-message (class:name ip:port)
</PRE>
or
<PRE>
  XPA$MESSAGE message 	  (class:name ip:port)
</PRE>
<P>
Note that when there is an error stored in an messages entry, the
corresponding bufs and lens entry may or may not be NULL and 0
(respectively), depending on the particularities of the server.

<P>
The return value will contain the actual number of servers that were
processed.  This value thus will hold the number of valid entries in
the bufs, lens, names, and messages arrays, and can be used to loop
through these arrays.  In names and/or messages is NULL, no information is
passed back in that array.

<P>
The mode string is of the form: "key1=value1,key2=value2,..."
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server (after callback completes)
</PRE>
<P>
The ack keyword is not very useful, since the server completes the callback
in order to return the data anyway.  It is here for completion (and perhaps
for future usefulness).
					   
<P>
<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;
  #define NXPA 10
  int  i, got;
  int fds[NXPA];
  char *names[NXPA];
  char *messages[NXPA];
  for(i=0; i&lt;NXPA; i++)
    fds[i] = open(...);
  got = XPAGetFd(NULL, "ds9", "file", NULL, fds, names, messages, NXPA);
  for(i=0; i&lt;got; i++){
    if( messages[i] != NULL ){
      /* error processing */
      fprintf(stderr, "ERROR: %s (%s)\n", messages[i], names[i]);
    }
    if( names[i] )
      free(names[i]);
    if( messages[i] )
      free(messages[i]);
  }
</PRE>

<!-- =defdoc xpasetfd xpasetfd 3 -->

<!-- =section xpasetfd NAME -->
<H2><A NAME="xpasetfd">XPASetFd: send data from stdin to one or more XPA servers</A></H2>

<!-- =section xpasetfd SYNOPSIS -->
</B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPASetFd(XPA xpa,
               char *template, char *paramlist, char *mode,
	       int fd, char **names, char **messages, int n)
</PRE>
</B>

<!-- =section xpasetfd DESCRIPTION -->
<P>
Read data from a standard I/O fd and send it to one or more XPA
servers whose class:name identifier matches the specified
<A HREF="./template.html">template.

<P>
A
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most n matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, then the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPASetFd() routine then reads bytes from the specified fd
until EOF and sends these bytes to the XPA servers.  
The final parameter n specifies the maximum number of servers to contact.
A string containing the class:name and ip:port of each server is returned in
the name array.  If a given server returned an error, then the error
message will be stored in the associated element of the messages array.
NB: if specified, the name and messages arrays must be of size n or greater.

<P>
The return value will contain the actual number of servers that were
processed.  This value thus will hold the number of valid entries in
the names and messages arrays, and can be used to loop through these
arrays.  In names and/or messages is NULL, no information is passed back
in that array.

<P>
The mode string is of the form: "key1=value1,key2=value2,..."
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server (after callback completes)
  verify	true/false	false		send buf from XPASet[Fd] to stdout
</PRE>
<P>
The ack keyword is useful in cases where one does not want to wait for
the server to complete, e.g. is a lot of processing needs to be done
on the passed data or when the success of the server operation is not
relevant to the client.

<P>
<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;

  #define NXPA 10
  int  i, got;
  int fd;
  char *names[NXPA];
  char *messages[NXPA];
  fd = open(...);
  got = XPASetFd(NULL, "ds9", "fits", NULL, fd, names, messages, NXPA);
  for(i=0; i&lt;got; i++){
    if( messages[i] != NULL ){
      /* error processing */
      fprintf(stderr, "ERROR: %s (%s)\n", messages[i], names[i]);
    }
    if( names[i] )
      free(names[i]);
    if( messages[i] )
      free(messages[i]);
  }
</PRE>

<!-- =defdoc xpaopen xpaopen 3 -->

<!-- =section xpaopen NAME -->
<H2><A NAME="xpaopen">XPAOpen: allocate a persistent client handle</A></H2>

<!-- =section xpaopen SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  XPA XPAOpen(char *mode);
</PRE>
</B>

<!-- =section xpaopen DESCRIPTION -->
<P>
XPAOpen() allocates a persistent XPA struct that can be used with
calls to XPAGet(), XPASet(), XPAInfo(), XPAGetFd(), and
XPASetFd(). Persistence means that a connection to an XPA server is
not closed when one of the above calls is completed but will be
re-used on successive calls. Using XPAOpen() therefore saves the time
it takes to connect to a server, which could be significant with slow
connections or if there will be a large number of exchanges with a
given access point.  The mode argument currently is ignored ("reserved
for future use").

<P>
An XPA struct is returned if XPAOpen() was successful; otherwise NULL
is returned. This returned struct can be passed as the first argument
to XPAGet(), etc.  Those calls will update the list of active XPA
connections.  Already connected servers (from a previous call) are
left connected and new servers also will be connected.  Old servers
(from a previous call) that are no longer needed are disconnected.
The connected servers will remain connected when the next call to
XPAGet() is made and connections are once again updated.

<P>
<B>Example:</B>
<PRE>
 #include &lt;xpa.h&gt;

  XPA xpa;
  xpa = XPAOpen(NULL);
</PRE>

<!-- =defdoc xpaclose xpaclose 3 -->

<!-- =section xpaclose NAME -->
<H2><A NAME="xpaclose">XPAClose: close a persistent XPA client handle</A></H2>

<!-- =section xpaclose SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  void XPAClose(XPA xpa);
</PRE>
</B>

<!-- =section xpaclose DESCRIPTION -->
<P>
XPAClose closes the persistent connections associated with this XPA struct
and frees all allocated space. It also closes the open sockets connections
to all XPA servers that were opened using this handle.

<P>
<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;

  XPA xpa;
  XPAClose(xpa);
</PRE>

<!-- =defdoc xpanslookup xpanslookup 3 -->

<!-- =section xpanslookup NAME -->
<H2><A NAME="xpanslookup">XPANSLookup: lookup registered XPA access points</A></H2>

<!-- =section xpanslookup SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPANSLookup(XPA xpa,
  	          char *template, char type,
	          char ***classes, char ***names,
	          char ***methods, char ***infos)
</PRE>
</B>

<!-- =section xpanslookup DESCRIPTION -->
<P>
XPA routines act on a class:name identifier in such a way
that all access points that match the identifier are processed.  It is
sometimes desirable to choose specific access points from the
candidates that match the
<A HREF="./template.html">template</A>.  In order to do this, the
XPANSLookup routine can be called to return a list of matches, so that
specific class:name instances can then be fed to XPAGet(), XPASet(), etc.

<P> The first argument is an optional XPA struct. If non-NULL, the
existing name server connection associated with the specified xpa is
used to query the xpans name server for matching templates. Otherwise,
a new (temporary) connection is established with the name server.

<P>
The second argument to XPANSLookup is the class:name 
<A HREF="./template.html">template</A>
to match.

<P>
The third argument for XPANSLookup() is the type of access and can be
any combination of:
<PRE>
  type   	explanation
  ------	-----------
  g		xpaget calls can be made on this access point
  s		xpaset calls can be made on this access point
  i		xpainfo calls can be made on this access point
</PRE>
<P>
The call typically specifies only one of these at a time.

<P>
The final arguments are pointers to arrays that will be filled
in and returned by the name server. The name server will allocate and
return arrays filled with the classes, names, and methods of all XPA
access points that match the <A HREF="./template.html">template</A>
and have the specified type. Also returned are info strings, which
generally are used internally by the client routines. These can be
ignored (but the strings must be freed).  The function returns the
number of matches. The returned value can be used to loop through the
matches:

<B>Example:</B>
<PRE>
  #include &lt;xpa.h&gt;

  char **classes;
  char **names;
  char **methods;
  char **infos;
  int i, n;
  n = XPANSLookup(NULL, "foo*", "g", &classes, &names, &methods, &infos);
  for(i=0; i&lt;n; i++){
    [more specific checks on possibilities ...]
    [perhaps a call to XPAGet for those that pass, etc. ...]
    /* don't forget to free alloc'ed strings when done */
    free(classes[i]);
    free(names[i]);
    free(methods[i]);
    free(infos[i]);
  }
  /* free up arrays alloc'ed by names server */
  if( n > 0 ){
    free(classes);
    free(names);
    free(methods);
    free(infos);
  }
</PRE>
<P>
The specified 
<A HREF="./template.html">template</A>
also can be a host:port specification, for example:
<PRE>
  myhost:12345
</PRE>
<P>
In this case, no connection is made to the name server. Instead, the
call will return one entry such that the ip array contains the ip for
the specified host and the port array contains the port.  The class
and name entries are set to the character "?", since the class and
name of the access point are not known. 

<!-- =defdoc xpaaccess xpaaccess 3 -->

<!-- =section xpaaccess NAME -->
<H2><A NAME="xpaaccess">XPAAccess: return XPA access points matching
template (XPA 2.1 and above)</A></H2>

<!-- =section xpaaccess SYNOPSIS -->
<B>
<PRE>
  #include &lt;xpa.h&gt;

  int XPAAccess(XPA xpa,
	        char *template, char *paramlist, char *mode,
	        char **names, char **messages, int n);
</PRE>
</B>

<!-- =section xpaaccess DESCRIPTION -->
<P>
The XPAAccess routine returns the public access points that match the
specified second argument <A HREF="./template.html">template</A> and
have the specified access type.

<P>
A
<A HREF="./template.html">template</A>
of the form "class1:name1" is sent to the
XPA name server, which returns a list of at most n matching XPA
servers.  A connection is established with each of these servers and
the paramlist string is passed to the server as the data transfer
request is initiated. If an XPA struct is passed to the call, then the
persistent connections are updated as described above. Otherwise,
temporary connections are made to the servers (which will be closed
when the call completes).

<P>
The XPAAccess() routine retrieves names from at most n XPA servers
that match the specified template and that were checked for access
using the specified mode.  The return string contains both the
class:name and ip:port.  If a given server returned an error or the
server callback sends a message back to the client, then the message
will be stored in the associated element of the messages array.
NB: if specified, the name and messages arrays must be of size n or greater.

<P>
The returned message string will be of the form:
<PRE>
  XPA$ERROR error-message (class:name ip:port)
</PRE>
<P>
Note that names of matching registered access points are always
returned but may not be valid; it is not sufficient to assume that the
returned number of access points is the number of valid access points.
Rather, it is essential to check the messages array for error
messages.  Any string in the messages array is an error message and
indicated that the associated access point is not available.

<P>
For example, assume that a server registers a number of access points
but delays entering its event loop. If a call to XPAAccess() is made
before the event loop is entered, the call will timeout (after waiting
for the long timeout period) and return an error of the form:
<PRE>
  XPA$ERROR: timeout waiting for server authentication (XPA:xpa1)
</PRE>
The error means that the XPA access point has been registered but is
not yet available (because events are not being processed). When the
server finally enters its event loop, subsequent calls to XPAAccess()
will return successfully.

<P>
NB: This routine only works with XPA servers built with XPA 2.1.x and later.
Servers with older versions of XPA will return the error message:

  XPA$ERROR invalid xpa command in initialization string

If you get this error message, then the old server actually is ready
for access, since it got to the point of fielding the query! The
xpaaccess program, for example, ignores this message in order to work
properly with older servers.

<P>
The third argument for XPAAccess() is the type of access and can be
any combination of:
<PRE>
  type   	explanation
  ------	-----------
  g		xpaget calls can be made on this access point
  s		xpaset calls can be made on this access point
  i		xpainfo calls can be made on this access point
</PRE>
<P>
The mode string argument is of the form: "key1=value1,key2=value2,..."
The following keywords are recognized:
<PRE>
  key   	value		default		explanation
  ------	--------	--------	-----------
  ack		true/false	true		if false, don't wait for ack from server (after callback completes)
</PRE>
<P>
The ack keyword is not very useful, since the server completes the callback
in order to return the data anyway.  It is here for completion (and perhaps
for future usefulness).

<!-- =section xpaclient SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpaget SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpaset SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpainfo SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpagetfd SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpasetfd SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpaopen SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpaclose SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpanslookup SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =section xpaaccess SEE ALSO -->
<!-- =text See xpa(n) for a list of XPA help pages -->
<!-- =stop -->

<P>
<A HREF="./help.html">Go to XPA Help Index</A>

<H5>Last updated: March 10, 2007</H5>

</BODY>
</HTML>