added option to configure a custom user agent within allip networks

defaults/yacy.init

@@ -1231,3 +1231,17 @@ greedylearning.active = true
 # postprocessing steering
 postprocessing.maximum_load = 2.5
 postprocessing.minimum_ram = 536870912
+
+# Custom user agents for 'allip' networks:
+# This user agent is only available if the network is set to 'allip' (which is a non-limited domain 'network'
+# without p2p options). Changing this will NOT change the default YaCy user agent, it will only provide an
+# agent which is available at crawl start within 'allip'. The userAgent.name is the identifier for the
+# robots.txt file which YaCy always obeys for the given name or a wildcard for robot types.
+# If any part of this custom user agent name or string includes the phrase 'yacy', it will be IGNORED
+# to prevent fraud, DoS or bad behavior in the name of YaCy.
+# To use this user agent option, you must define completely different names and strings
+# and remove the given example here, which will be ignored by default.
+crawler.userAgent.name = yacybot
+crawler.userAgent.string = yacybot ($$SYSTEM$$) http://yacy.net/bot.html
+crawler.userAgent.minimumdelta = 500
+crawler.userAgent.clienttimeout = 10000

htroot/CrawlStartExpert.java

@@ -494,6 +494,7 @@ public class CrawlStartExpert {
             agentNames.add(ClientIdentification.googleAgentName);
             if (sb.isAllIPMode()) {
                 agentNames.add(ClientIdentification.browserAgentName);
+                if (ClientIdentification.getAgent(ClientIdentification.customAgentName) != null) agentNames.add(ClientIdentification.customAgentName);
             }
             String defaultAgentName = agentNames.get(0);
             if (post != null && post.containsKey("agentName")) {

htroot/Crawler_p.java

@@ -319,8 +319,10 @@ public class Crawler_p {
 
                 env.setConfig("storeHTCache", storeHTCache);
                 
-                String agentName = post.get("agentName", sb.isIntranetMode() ? ClientIdentification.yacyIntranetCrawlerAgentName : ClientIdentification.yacyInternetCrawlerAgentName);
+                String defaultAgentName = sb.isIntranetMode() ? ClientIdentification.yacyIntranetCrawlerAgentName : ClientIdentification.yacyInternetCrawlerAgentName;
+                String agentName = post.get("agentName", defaultAgentName);
                 ClientIdentification.Agent agent = ClientIdentification.getAgent(agentName);
+                if (agent == null) agent = ClientIdentification.getAgent(defaultAgentName);
 
                 CacheStrategy cachePolicy = CacheStrategy.parse(post.get("cachePolicy", "iffresh"));
                 if (cachePolicy == null) cachePolicy = CacheStrategy.IFFRESH;

source/net/yacy/cora/protocol/ClientIdentification.java

@@ -77,6 +77,7 @@ public class ClientIdentification {
     public final static Agent browserAgent = new Agent(browserAgents[random.nextInt(browserAgents.length)], new String[]{"Mozilla"}, minimumLocalDeltaInit, clientTimeoutInit);
     public final static String yacyProxyAgentName = "YaCyProxy";
     public final static Agent yacyProxyAgent = new Agent("yacy - this is a proxy access through YaCy from a browser, not a robot (the yacy bot user agent is 'yacybot')", new String[]{"yacy"}, minimumGlobalDeltaInit, clientTimeoutInit);
+    public final static String customAgentName = "Custom Agent";
 
     static {
         generateYaCyBot("new");
@@ -105,6 +106,12 @@ public class ClientIdentification {
         agents.put(yacyIntranetCrawlerAgentName, yacyIntranetCrawlerAgent);
     }
     
+    public static void generateCustomBot(String name, String string, int minimumdelta, int clienttimeout) {
+        if (name.toLowerCase().indexOf("yacy") >= 0 || string.toLowerCase().indexOf("yacy") >= 0) return; // don't allow 'yacy' in custom bot strings
+        String agentString = string.replace("$$SYSTEM$$", yacySystem.replace("java", "O"));
+        agents.put(customAgentName, new Agent(agentString, new String[]{name}, minimumdelta, clienttimeout));
+    }
+
     /**
      * get the default agent
      * @param newagent

source/net/yacy/search/Switchboard.java

@@ -395,12 +395,18 @@ public final class Switchboard extends serverSwitch {
 
         // load the network definition
         try {
-            overwriteNetworkDefinition();
+            overwriteNetworkDefinition(getSysinfo());
         } catch (final FileNotFoundException e) {
             ConcurrentLog.logException(e);
         } catch (final IOException e) {
             ConcurrentLog.logException(e);
         }
+        // create custom user agent
+        ClientIdentification.generateCustomBot(
+                getConfig(SwitchboardConstants.CRAWLER_USER_AGENT_NAME, ""),
+                getConfig(SwitchboardConstants.CRAWLER_USER_AGENT_STRING, ""),
+                (int) getConfigLong(SwitchboardConstants.CRAWLER_USER_AGENT_MINIMUMDELTA, 500),
+                (int) getConfigLong(SwitchboardConstants.CRAWLER_USER_AGENT_CLIENTTIMEOUT , 1000));
         
         // start indexing management
         this.log.config("Starting Indexing Management");
@@ -1140,6 +1146,10 @@ public final class Switchboard extends serverSwitch {
         this.log.config("Finished Switchboard Initialization");
     }
 
+    final String getSysinfo() {
+        return getConfig(SwitchboardConstants.NETWORK_NAME, "") + (isRobinsonMode() ? "-" : "/") + getConfig(SwitchboardConstants.NETWORK_DOMAIN, "global");
+    }
+    
     @Override
     public void setHttpServer(YaCyHttpServer server) {
         super.setHttpServer(server);
@@ -1162,7 +1172,7 @@ public final class Switchboard extends serverSwitch {
             + this.indexingStorageProcessor.getQueueSize();
     }
 
-    public void overwriteNetworkDefinition() throws FileNotFoundException, IOException {
+    public void overwriteNetworkDefinition(final String sysinfo) throws FileNotFoundException, IOException {
 
         // load network configuration into settings
         String networkUnitDefinition =
@@ -1264,9 +1274,7 @@ public final class Switchboard extends serverSwitch {
         }
         */
         // write the YaCy network identification inside the yacybot client user agent to distinguish networks
-        ClientIdentification.generateYaCyBot(getConfig(SwitchboardConstants.NETWORK_NAME, "")
-                + (isRobinsonMode() ? "-" : "/")
-                + getConfig(SwitchboardConstants.NETWORK_DOMAIN, "global"));
+        ClientIdentification.generateYaCyBot(sysinfo);
     }
 
     public void switchNetwork(final String networkDefinition) throws FileNotFoundException, IOException {
@@ -1309,7 +1317,7 @@ public final class Switchboard extends serverSwitch {
 
             // new properties
             setConfig("network.unit.definition", networkDefinition);
-            overwriteNetworkDefinition();
+            overwriteNetworkDefinition(getSysinfo());
             final File indexPrimaryPath =
                 getDataPath(SwitchboardConstants.INDEX_PRIMARY_PATH, SwitchboardConstants.INDEX_PATH_DEFAULT);
             final int wordCacheMaxCount =

source/net/yacy/search/SwitchboardConstants.java

@@ -340,6 +340,10 @@ public final class SwitchboardConstants {
     public static final String CRAWLER_FOLLOW_REDIRECTS         = "crawler.http.FollowRedirects"; // ignore the target url and follow to the redirect
     public static final String CRAWLER_RECORD_REDIRECTS         = "crawler.http.RecordRedirects"; // record the ignored redirected page to the index store
     
+    public static final String CRAWLER_USER_AGENT_NAME          = "crawler.userAgent.name";
+    public static final String CRAWLER_USER_AGENT_STRING        = "crawler.userAgent.string";
+    public static final String CRAWLER_USER_AGENT_MINIMUMDELTA  = "crawler.userAgent.minimumdelta";
+    public static final String CRAWLER_USER_AGENT_CLIENTTIMEOUT = "crawler.userAgent.clienttimeout";
     
     /**
      * debug flags