Dominic Cronin's questions

We have a problem with spelling suggestions from our Google search appliance. For a particular search term, it offers a completely incorrect spelling suggestion, which, apart from being unhelpful, is at odds with the otherwise carefully managed tone of our web site. The word is present in some documents, but is completely unrelated to the search term.

From what I have read so far, spelling suggestions are not something that you can explicitly manage, but are gathered from within the corpus the GSA is indexing. We can see that it's not a default result, because the index of our development systems doesn't offer this suggestion.

The Google documentation says: "You cannot edit the search appliance's spelling dictionary." It then goes on to list some other options for improving searches, none of which seems to include being able to remove or prevent such spurious suggestions.

So what are the options? Can we directly influence the spelling suggestions for a specific word or words in some other way? Can we get the machine-learning system to forget what it's learned so far? Are there other possible solutions?

I am setting up a local instance of the W3c validator on a computer running Centos 6.4. I have installed it using yum. For HTML5 validation, the validator hands off to a separate implementation, available from validator.nu

I can run the validator.nu implementation succesfully under Jetty, using the script provided with the distribution. I can test this successfully by pointing a browser directly at localhost:8888 and using it to validate a web site.

When I try to use my w3c validator to validate an HTML5 site, it attempts the hand-off and fails with the following message:

Checking the Document Type of this document requires the help of an external tool which was either not enabled in this validator, or is currently unavailable. Check in the validator's system configuration that HTML5 Validator is enabled and functional. The error encountered was: 500 Can't connect to localhost:8888 (connect: Permission denied)

Disabling the firewall does not help. Neither apache (which hosts the w3c validator) nor Jetty (which hosts the validator.nu validator) appear to log anything relevant.

What relevant differences might there be between hitting localhost:8888 from a browser and from a cgi-bin running under apache?

Update: /var/log/audit/audit/log has the following:

type=AVC msg=audit(1393503835.221:31235): avc: denied { name_connect } for pid=40041 comm="check" dest=8888 scontext=unconfined_u:system_r:httpd_w3c_validator_script_t:s0 tcontext=system_u:object_r:port_t:s0 tclass=tcp_socket type=SYSCALL msg=audit(1393503835.221:31235): arch=c000003e syscall=42 success=no exit=-13 a0=3 a1=2340db0 a2=10 a3=21936e0 items=0 ppid=36416 pid=40041 auid=500 uid=48 gid=48 euid=48 suid=48 fsuid=48 egid=48 sgid=48 fsgid=48 tty=(none) ses=1 comm="check" exe="/usr/bin/perl" subj=unconfined_u:system_r:httpd_w3c_validator_script_t:s0 key=(null)

I am using Microsoft Search Server 2010 Express to spider a web site and provide search facilities to the site users. In the html of the site, we have various elements that are hidden by means of a css class. For example:

<p class="hidden">Go to <a href="#mainNav" title="main menu">main menu</a></p>

Such text is intended to provide additional assistance for the visually impaired, and is not meaningful in the context of search results.

How can I configure Search Server so that it doesn't include text like this in the hit context of a result?

I have created a dynamic component presentation which contains UGC TCDL tags. The DCP is published to the broker database as REL.

I can see in the COMPONENT_PRESESENTATIONS table that the tags are present. (<tcdl:PostRating .../>, <tcdl:ContextVariable..../> and the like.)

In the COMPONENT_PRES_META_DATA table, I can see that the DCP in question has a COMPONENT_OUTPUT_FORMAT of REL.

I have reduced my test case to a page that does simply the following:

<%@ Page Language="C#" %>
<%
  var ass = new Tridion.ContentDelivery.DynamicContent.ComponentPresentationAssembler("tcm:19-448-64");
  Response.Write(ass.GetContent("tcm:19-429", "tcm:19-449-32"));
%>

This page renders the content of the DCP exactly as it is to be found in the database, with the <tcdl:/> tags still intact.

In the bin\config\cd_dynamic_conf.xml file in the web site, is configured a Configuration/TCDLEngine/Renderer/TagBundle element, as follows:

<TagBundle Resource="ugc_renderer_bundle.xml"/>

and bin\lib\ugc_dynamic.jar is present

Where else should I be looking for problems with my configuration? Are there any other possible causes for this problem, or approaches I might take?

I have implemented a Spam filter based on the Tridion UGC API, and I now wish to configure my UGC web service to make use of it. So far the spam filter appears not to be invoked when adding a comment via the service. My current configuration is as follows:

I have placed the assembly DLL in the bin directory of the web service application. I have edited the web.config file to add the following:

<appSettings>
    <add key="Comment.SpamFilter.Implementation" value="UGCDemo.BadWordsFilter"/>
    <add key="Comment.SpamFilter.Assembly" value="SpamFilter.dll"/>
</appSettings

I am able to verify the name of the implementation class and of the assembly, as you can see by the following PowerShell output:

PS C:\inetpub\LiveCDService
> ls

    Directory: C:\inetpub\LiveCDService


Mode                LastWriteTime     Length Name
----                -------------     ------ ----
d----         5/28/2012   5:05 PM            bin
-a---         11/4/2011   3:38 PM        212 access_token.svc
-a---         11/4/2011   3:38 PM        323 Global.asax
-a---         11/4/2011   3:38 PM        138 linking.svc
-a---         11/4/2011   3:38 PM        198 odata.svc
-a---         5/28/2012   5:41 PM      11073 Web.config
-a---         5/28/2012   5:10 PM      11073 Web.config~


PS C:\inetpub\LiveCDService
> add-type -path .\bin\SpamFilter.dll
PS C:\inetpub\LiveCDService
> $foo = new-object UGCDemo.BadWordsFilter
PS C:\inetpub\LiveCDService
> $foo | gm


   TypeName: UGCDemo.BadWordsFilter

Name            MemberType Definition
----            ---------- ----------
Equals          Method     bool Equals(System.Object obj)
GetHashCode     Method     int GetHashCode()
GetType         Method     type GetType()
ToString        Method     string ToString()
ValidateComment Method     Tridion.ContentDelivery.UGC.Web.Model.Comment Valid

I have the logging in debug mode, and I can't see any output that might relate to loading a spam filter assembly or anything similar. What other configuration changes might I need to get this functionality working? What other tests might I apply to discover more about the problem?

I have installed the Tridion content delivery web service on a 64 bit Windows Server 2008 R2 system, intending to configure UGC. (In fact I have two services, which both exhibit the same problem).

My service has the ambient framework HttpModule configured, and the cd_ambient_conf.xml is a copy of the default version shipped with the product, however the Security element is commented out. I am testing by visiting the odata.svc endpoint in the browser.

In the Cartridges element, I have the following:

<Cartridges>
        <Cartridge File="cd_ambient_cartridge_conf.xml"/>
        <Cartridge File="ugc_ambient_cartridge.xml"/> 
</Cartridges>

When configured like this, or if I comment out the cd_ambient cartridge and leave the ugc_ambient cartridge in, I get a null reference exception:

[NullReferenceException: Object reference not set to an instance of an object.]
Tridion.ContentDelivery.AmbientData.HttpModule.OnRequestStart
                                                (Object sender, EventArgs e) +292
System.Web.SyncEventExecutionStep.System.Web
                                    .HttpApplication.IExecutionStep.Execute() +79
System.Web.HttpApplication.ExecuteStep(
                       IExecutionStep step, Boolean& completedSynchronously) +269

If I configure only the cd_ambient cartridge, then I get the following error:

[AmbientDataException: Error while configuring ambient data framework]

If I remove both cartridges, the service comes up without an error. I'm speculating that the null reference errors are caused by perhaps a missing configuration value, but at least cd_ambient_conf.xml passes schema validation, so perhaps not.

So... what is likely to be the cause of this problem, and are there any debugging or analysis techniques that may help me to find out more?

EDIT: I have now removed the cd_ambient_cartridge_conf.xml reference from within Cartridges. So with only the ugc_ambient_cartridge.xml reference, I get the following error in the cd_core log (and I'm back to getting the [AmbientDataException: Error while configuring ambient data framework] error):

2012-05-11 15:03:51,073 WARN  AmbientDataContext - There is no current ambient 
       data context - the ambient data framework is not properly initialised
2012-05-11 15:03:51,171 ERROR EngineFactory - Unable to configure the 
                                                       AmbientData Framework
com.tridion.configuration.ConfigurationException: 
   Can't find configuration file: [ ugc_ambient_cartridge.xml ]
at         com.tridion.configuration.XMLConfigurationReader.readConfiguration(
                                XMLConfigurationReader.java:92) ~[cd_core.jar:na]
at         com.tridion.ambientdata.AmbientDataConfig.getCartridgeConfigurations(
                                 AmbientDataConfig.java:155) ~[cd_ambient.jar:na]
at com.tridion.ambientdata.EngineFactory.newEngine(
                                      EngineFactory.java:45) ~[cd_ambient.jar:na]

I have recently compiled a new kernel for my linux server. It is almost identical with the previous kernel (I added cifs support). Because physical access to the server is limited, I configured grub to fallback to the old kernel if the new kernel failed to boot cleanly. I can now boot the server, but I don't know which kernel is running.

I tried doing "uname -a" and it gives me the kernel name followed by #2. Is the #2 relevant to my problem?

How can I find out which of the two possible kernels is running? Ideally I'd like to relate it to the order listed in my grub.conf