adelton

Docserver -- installation steps

The steps to do and check for are:

  1. Have Windows server with Perl and Word/Excel.
  2. Install additional Perl modules.
  3. Have client machine with Perl.
  4. Check that the versions of Storable module are compatible.
  5. Check that the Doc(client|server)/Config.pm files are reasonable.
  6. Run tests locally on the server.
  7. Debug errors you might get from running the tests.
  8. Run remote tests from the client.
  9. Install and setup the docserver as service.

This document was back-translated from Czech original; it is possible that names of some options/services on the MS side do not match exactly -- please use common sense.

Windows server

For the server, we use WinNT with 630 build of ActiveStave Perl; previous builds worked fine as well. Different version of Perl or different version of OS may/should work for you but please note that I won't be of much help resolving OS problems on this side.

Install Perl and additional modules

After installing the ActiveState Perl, use ppm to further install RPC::PlServer, Win32::OLE and WIN32::API packages, if they didn't come with the default ActiveState distribution. Run ppm and on its prompt enter install RPC::PlServer <Enter> and similarily for those other two packages. You also need Storable module (from those that are rare) but that seems to be installed by ActiveState by default when installing RPC::PlServer.

If you experience problems with the repository at activestate.com because of differing versions of Perl, run ppm genconfig and edit the output to match the correct locations on activestate.com. If you used some other Perl distribution, check documentation that came with that distribution for information about installing these modules.

Install Word and Excel

You need to have Word (and Excel for xls conversion) installed, the newer version the better. The 2000 versions seem to be working fine. Install converters Word/Text with Layout and Converters/Text/RTF and export to Word 6.0. If you plan to do some development or debugging of Docserver, install Help and Tools for Office/Help for Visual Basic. It may be necessary to set user's name and initials to space, because of an error in one convertor. You can also consider installing applicable patches from http://office.microsoft.com.

In Word and Excel in Options/Macro/Security select High securify level for the user under which the Docserver will be running, and remove any trusted resources and setup no trust to any plugins, etc.

The following steps will be shown on the SYSTEM account. You can change the security leve in registry by setting HKU/.default/SW/MS/Office/9.0/Word/Security/Level to 3 (dword) and DontTrustInstalledFiles to 1 (dword). For Excel it is similar, the registry records are 

[HKEY_USERS\.default\Software\Microsoft\Office\9.0\Word\Security]
"Level"=dword:00000003
"DontTrustInstalledFiles"=dword:00000001

[HKEY_USERS\.default\Software\Microsoft\Office\9.0\Excel\Security]
"Level"=dword:00000003
"DontTrustInstalledFiles"=dword:00000001

I don't know how to add a trusted resource, well I do not know how to do that without running Word -- the safest would be to run under SYSTEM account something like 

use Win32::OLE;
my $word = Win32::OLE->new('Word.Application');
$word->{'Visible'} = 1;

and do it manually.

Setup PostScript printers

For conversions to PostScript, install printer with Adobe PS Printer Driver (http://www.adobe.com/products/printerdrivers/main.html) as local on port FILE: and with driver Adobe Generic PS. You may want to setup ASCII output protokol, no ^D before nor after the job, and check that the output format matches your local customs (A4 for Europe, probably). Set the memory for PS to value reasonable for your needs, 4096kB works fine for us. In document defaults set TrueType Font to Download as Softfont, TrueType Font Download Option to Outline and PostScript Output Option to Optimize for Speed. It is recommended to install another printer (ps1 in Docserver) as Optimize for Portability (that gives you reasonable PostScript that you can work with in Un*x). Yet another printer (and another setting in Docserver/Config.pm) could be with Substitute with Device Fonts Automatic and Optimize for Speed for smallest output PostScript, however if you're in non-English speaking country, your diacritical characters will be of worse quality. If you decide to use some other driver (for example HP), set Page Independence Yes and Generate Job Control Code No.

Client machine

For the client, any machine is OK, you need the RPC::PlClient package here. The version of Storable module has to be the same on the client and on the server (or they have to be close/compatible -- 1.0.13 on client and 1.0.10 on Windows server seem to work). So you will probably want to install whatever is available for the ActiveState on Windows and then match the version on the client (Un*x) side.

Installing the docserver/docclient Perl code -- it goes the ordinary way 

perl Makefile.PL
make
make test
make install

Installing Docserver.pm on Un*x machine doesn't hurt -- it just won't be used.

Many people don't have working make on Windows, so alternatively you can do 

perl Makefile.PL
perl do_test.pl

and copy lib/ and bin/ directories to wherever you like them.

Configuration is stored in lib/Docclient/Config.pm for client side and lib/Docserver/Config.pm for the server. You should match the machine names and the port numbers. On the server, check that the printer names for PostScript output (ps, ps1) match the names of PostScript printers (drivers) in your system.

On one system we tested, Excel under SYSTEM had troubles using printers installed under different account. It can be solved by copying HKEY_USER\[user]\Printers and HKEY_USER\[user]\Software\Microsoft\Windows NT\CurrentVersion\[Devices|PrinterPorts|Windows\Device] to appropriate places in HKEY_USER\.default.

Running docserver as service

To setup docserver as a service, you can try FireDaemon from http://www.firedaemon.com/. After installation, follow documentation and set variables MXBIN and MXHOME, and then run (this should be one line) 

firedaemon -i docserver working\directory path\to\perl.exe "[-Ipath] [path\to\]docserver.pl" N 0 0 Y Y

With this setup you won't be able to log the output of docserver, you can try something like (again, one line) 

firedaemon -i docserver working\directory path\to\cmd.exe "/c docserver.cmd" N 0 0 Y Y

where in docserver.cmd you'll have (two lines, this time) 

start /b [path\]cmd.exe /c ([path\]perl.exe [-Ilib] [path\]docserver.pl >>[path\]docserver.log 2>&1)
net stop docserver

If you setup the SYSTEM account in Control Panel/Services as the account under which docserver should run, you won't be able to kill the process (docserver.pid) from Task Manager, should it hang. You can use free pskill from psutils package from www.sysinternals.com, or it is possible to use FireDaemon to create service which would be checking and killing the process which got out of control.

Before restarting (after kill or crash), you have to do net stop docserver. If you use the docserver.cmd above, you can do something like 

(for /f %i in ([path\]docserver.pid) do [path\]pskill %i) & net start docserver

You can also use AutoExNT from Resource Kit. You need files autoexnt.exe, autoexnt.bat, instexnt.exe and servmess.dll. Except instexnt.exe they have to be copied to %systemroot%\system32. Autoexnt.bat should contain line 

start "docserver.pl" /DC:\directory\to\docserver.pl /MIN C:\full\path\to\perl.exe docserver.pl

The enter instexnt install /interactive and check Control Panels/ Services/AutoExNT/Startup for Automatic. You can start docserver as service without reboot using 

net start autoexnt

You can also try to use srvany from Resource Kit.

You should also be able to achieve the same result by directly editing registry.

Known bugs & limitations (not on our side)

The normal text convertor (txt1) puts header under the normal text.

When converting from xls95 to csv, convertor is adding extra sheet with the content of the first sheet after each macro.

The MS products may fail on you at any time under any conditions. This is not a 100 percent solution -- remember, the conversion from proprietary format is done by proprietary software. Hopefully you'll get reasonable results for most of your documents.