(crawler) Further rearrange crawl order

Limit crawl order preferrence to edu domains, to avoid hitting stuff like medium and wordpress with shotgun requests.

build.gradle

@@ -43,12 +43,11 @@ subprojects.forEach {it ->
 }
 
 ext {
-    jvmVersion=23
-    dockerImageBase='container-registry.oracle.com/graalvm/jdk:23'
+    jvmVersion = 24
+    dockerImageBase='container-registry.oracle.com/graalvm/jdk:24'
     dockerImageTag='latest'
     dockerImageRegistry='marginalia'
     jibVersion = '3.4.4'
-
 }
 
 idea {

code/common/db/java/nu/marginalia/db/DbDomainQueries.java

@@ -22,6 +22,7 @@ public class DbDomainQueries {
     private static final Logger logger = LoggerFactory.getLogger(DbDomainQueries.class);
 
     private final Cache<EdgeDomain, Integer> domainIdCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
+    private final Cache<EdgeDomain, DomainIdWithNode> domainWithNodeCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
     private final Cache<Integer, EdgeDomain> domainNameCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
     private final Cache<String, List<DomainWithNode>> siblingsCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
 
@@ -59,6 +60,34 @@ public class DbDomainQueries {
         }
     }
 
+
+    public DomainIdWithNode getDomainIdWithNode(EdgeDomain domain) throws NoSuchElementException {
+        try {
+            return domainWithNodeCache.get(domain, () -> {
+                try (var connection = dataSource.getConnection();
+                     var stmt = connection.prepareStatement("SELECT ID, NODE_AFFINITY FROM EC_DOMAIN WHERE DOMAIN_NAME=?")) {
+
+                    stmt.setString(1, domain.toString());
+                    var rsp = stmt.executeQuery();
+                    if (rsp.next()) {
+                        return new DomainIdWithNode(rsp.getInt(1), rsp.getInt(2));
+                    }
+                }
+                catch (SQLException ex) {
+                    throw new RuntimeException(ex);
+                }
+
+                throw new NoSuchElementException();
+            });
+        }
+        catch (UncheckedExecutionException ex) {
+            throw new NoSuchElementException();
+        }
+        catch (ExecutionException ex) {
+            throw new RuntimeException(ex.getCause());
+        }
+    }
+
     public OptionalInt tryGetDomainId(EdgeDomain domain) {
 
         Integer maybeId = domainIdCache.getIfPresent(domain);
@@ -145,4 +174,6 @@ public class DbDomainQueries {
             return nodeAffinity > 0;
         }
     }
+
+    public record DomainIdWithNode (int domainId, int nodeAffinity) { }
 }

code/common/model/java/nu/marginalia/model/EdgeDomain.java

@@ -14,7 +14,7 @@ public class EdgeDomain implements Serializable {
     @Nonnull
     public final String topDomain;
 
-    public EdgeDomain(String host) {
+    public EdgeDomain(@Nonnull String host) {
         Objects.requireNonNull(host, "domain name must not be null");
 
         host = host.toLowerCase();
@@ -61,6 +61,10 @@ public class EdgeDomain implements Serializable {
         this.topDomain = topDomain;
     }
 
+    public static String getTopDomain(String host) {
+        return new EdgeDomain(host).topDomain;
+    }
+
     private boolean looksLikeGovTld(String host) {
         if (host.length() < 8)
             return false;
@@ -116,24 +120,6 @@ public class EdgeDomain implements Serializable {
         return topDomain.substring(0, cutPoint).toLowerCase();
     }
 
-    public String getLongDomainKey() {
-        StringBuilder ret = new StringBuilder();
-
-        int cutPoint = topDomain.indexOf('.');
-        if (cutPoint < 0) {
-            ret.append(topDomain);
-        } else {
-            ret.append(topDomain, 0, cutPoint);
-        }
-
-        if (!subDomain.isEmpty() && !"www".equals(subDomain)) {
-            ret.append(":");
-            ret.append(subDomain);
-        }
-
-        return ret.toString().toLowerCase();
-    }
-
     /** If possible, try to provide an alias domain,
      * i.e. a domain name that is very likely to link to this one
      * */

code/common/service/test/nu/marginalia/service/discovery/ZkServiceRegistryTest.java

@@ -25,7 +25,7 @@ import static org.mockito.Mockito.when;
 class ZkServiceRegistryTest {
     private static final int ZOOKEEPER_PORT = 2181;
     private static final GenericContainer<?> zookeeper =
-            new GenericContainer<>("zookeeper:3.8.0")
+            new GenericContainer<>("zookeeper:3.8")
                     .withExposedPorts(ZOOKEEPER_PORT);
 
     List<ZkServiceRegistry> registries = new ArrayList<>();

code/functions/favicon/api/build.gradle

@@ -0,0 +1,47 @@
+plugins {
+    id 'java'
+
+    id "com.google.protobuf" version "0.9.4"
+    id 'jvm-test-suite'
+}
+
+java {
+    toolchain {
+        languageVersion.set(JavaLanguageVersion.of(rootProject.ext.jvmVersion))
+    }
+}
+
+jar.archiveBaseName = 'favicon-api'
+
+apply from: "$rootProject.projectDir/protobuf.gradle"
+apply from: "$rootProject.projectDir/srcsets.gradle"
+
+dependencies {
+    implementation project(':code:common:model')
+    implementation project(':code:common:config')
+    implementation project(':code:common:service')
+
+    implementation libs.bundles.slf4j
+
+    implementation libs.prometheus
+    implementation libs.notnull
+    implementation libs.guava
+    implementation dependencies.create(libs.guice.get()) {
+        exclude group: 'com.google.guava'
+    }
+    implementation libs.gson
+    implementation libs.bundles.protobuf
+    implementation libs.guava
+    libs.bundles.grpc.get().each {
+        implementation dependencies.create(it) {
+            exclude group: 'com.google.guava'
+        }
+    }
+
+
+
+    testImplementation libs.bundles.slf4j.test
+    testImplementation libs.bundles.junit
+    testImplementation libs.mockito
+
+}

code/functions/favicon/api/java/nu/marginalia/api/favicon/FaviconClient.java

@@ -0,0 +1,39 @@
+package nu.marginalia.api.favicon;
+
+import com.google.inject.Inject;
+import nu.marginalia.service.client.GrpcChannelPoolFactory;
+import nu.marginalia.service.client.GrpcMultiNodeChannelPool;
+import nu.marginalia.service.discovery.property.ServiceKey;
+import nu.marginalia.service.discovery.property.ServicePartition;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.util.Optional;
+
+public class FaviconClient {
+    private static final Logger logger = LoggerFactory.getLogger(FaviconClient.class);
+
+    private final GrpcMultiNodeChannelPool<FaviconAPIGrpc.FaviconAPIBlockingStub> channelPool;
+
+    @Inject
+    public FaviconClient(GrpcChannelPoolFactory factory) {
+        this.channelPool = factory.createMulti(
+                ServiceKey.forGrpcApi(FaviconAPIGrpc.class, ServicePartition.multi()),
+                FaviconAPIGrpc::newBlockingStub);
+    }
+
+    public record FaviconData(byte[] bytes, String contentType) {}
+
+
+    public Optional<FaviconData> getFavicon(String domain, int node) {
+        RpcFaviconResponse rsp = channelPool.call(FaviconAPIGrpc.FaviconAPIBlockingStub::getFavicon)
+                .forNode(node)
+                .run(RpcFaviconRequest.newBuilder().setDomain(domain).build());
+
+        if (rsp.getData().isEmpty())
+            return Optional.empty();
+
+        return Optional.of(new FaviconData(rsp.getData().toByteArray(), rsp.getContentType()));
+    }
+
+}

code/functions/favicon/api/src/main/protobuf/favicon.proto

@@ -0,0 +1,20 @@
+syntax="proto3";
+package marginalia.api.favicon;
+
+option java_package="nu.marginalia.api.favicon";
+option java_multiple_files=true;
+
+service FaviconAPI {
+  /** Fetches information about a domain. */
+  rpc getFavicon(RpcFaviconRequest) returns (RpcFaviconResponse) {}
+}
+
+message RpcFaviconRequest {
+  string domain = 1;
+}
+
+message RpcFaviconResponse {
+  string domain = 1;
+  bytes data = 2;
+  string contentType = 3;
+}

code/functions/favicon/build.gradle

@@ -0,0 +1,49 @@
+plugins {
+    id 'java'
+
+    id 'application'
+    id 'jvm-test-suite'
+}
+
+java {
+    toolchain {
+        languageVersion.set(JavaLanguageVersion.of(rootProject.ext.jvmVersion))
+    }
+}
+
+apply from: "$rootProject.projectDir/srcsets.gradle"
+
+dependencies {
+    implementation project(':code:common:config')
+    implementation project(':code:common:service')
+    implementation project(':code:common:model')
+    implementation project(':code:common:db')
+    implementation project(':code:functions:favicon:api')
+    implementation project(':code:processes:crawling-process')
+
+    implementation libs.bundles.slf4j
+
+    implementation libs.prometheus
+    implementation libs.guava
+    libs.bundles.grpc.get().each {
+        implementation dependencies.create(it) {
+            exclude group: 'com.google.guava'
+        }
+    }
+
+
+    implementation libs.notnull
+    implementation libs.guava
+    implementation dependencies.create(libs.guice.get()) {
+        exclude group: 'com.google.guava'
+    }
+    implementation dependencies.create(libs.spark.get()) {
+        exclude group: 'org.eclipse.jetty'
+    }
+
+    testImplementation libs.bundles.slf4j.test
+    testImplementation libs.bundles.junit
+    testImplementation libs.mockito
+
+
+}

code/functions/favicon/java/nu/marginalia/functions/favicon/FaviconGrpcService.java

@@ -0,0 +1,48 @@
+package nu.marginalia.functions.favicon;
+
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
+import com.google.protobuf.ByteString;
+import io.grpc.stub.StreamObserver;
+import nu.marginalia.api.favicon.FaviconAPIGrpc;
+import nu.marginalia.api.favicon.RpcFaviconRequest;
+import nu.marginalia.api.favicon.RpcFaviconResponse;
+import nu.marginalia.crawl.DomainStateDb;
+import nu.marginalia.service.server.DiscoverableService;
+
+import java.util.Optional;
+
+@Singleton
+public class FaviconGrpcService extends FaviconAPIGrpc.FaviconAPIImplBase implements DiscoverableService {
+    private final DomainStateDb domainStateDb;
+
+    @Inject
+    public FaviconGrpcService(DomainStateDb domainStateDb) {
+        this.domainStateDb = domainStateDb;
+    }
+
+    public boolean shouldRegisterService() {
+        return domainStateDb.isAvailable();
+    }
+
+    @Override
+    public void getFavicon(RpcFaviconRequest request, StreamObserver<RpcFaviconResponse> responseObserver) {
+        Optional<DomainStateDb.FaviconRecord> icon = domainStateDb.getIcon(request.getDomain());
+
+        RpcFaviconResponse response;
+        if (icon.isEmpty()) {
+            response = RpcFaviconResponse.newBuilder().build();
+        }
+        else {
+            var iconRecord = icon.get();
+            response = RpcFaviconResponse.newBuilder()
+                            .setContentType(iconRecord.contentType())
+                            .setDomain(request.getDomain())
+                            .setData(ByteString.copyFrom(iconRecord.imageData()))
+                            .build();
+        }
+
+        responseObserver.onNext(response);
+        responseObserver.onCompleted();
+    }
+}

code/libraries/blocking-thread-pool/java/nu/marginalia/util/SimpleBlockingThreadPool.java

@@ -23,16 +23,33 @@ public class SimpleBlockingThreadPool {
     private final Logger logger = LoggerFactory.getLogger(SimpleBlockingThreadPool.class);
 
     public SimpleBlockingThreadPool(String name, int poolSize, int queueSize) {
+        this(name, poolSize, queueSize, ThreadType.PLATFORM);
+    }
+
+    public SimpleBlockingThreadPool(String name, int poolSize, int queueSize, ThreadType threadType) {
         tasks = new ArrayBlockingQueue<>(queueSize);
 
         for (int i = 0; i < poolSize; i++) {
-            Thread worker = new Thread(this::worker, name  + "[" + i + "]");
-            worker.setDaemon(true);
-            worker.start();
+
+            Thread.Builder threadBuilder = switch (threadType) {
+                case VIRTUAL -> Thread.ofVirtual();
+                case PLATFORM -> Thread.ofPlatform().daemon(true);
+            };
+
+            Thread worker = threadBuilder
+                    .name(name  + "[" + i + "]")
+                    .start(this::worker);
+
             workers.add(worker);
         }
 
     }
+
+    public enum ThreadType {
+        VIRTUAL,
+        PLATFORM
+    }
+
     public void submit(Task task) throws InterruptedException {
         tasks.put(task);
     }

code/processes/crawling-process/java/nu/marginalia/crawl/CrawlerMain.java

@@ -105,7 +105,8 @@ public class CrawlerMain extends ProcessMainClass {
 
         pool = new SimpleBlockingThreadPool("CrawlerPool",
                 Integer.getInteger("crawler.poolSize", 256),
-                1);
+                1,
+                SimpleBlockingThreadPool.ThreadType.VIRTUAL);
 
 
         // Wait for the blacklist to be loaded before starting the crawl
@@ -221,10 +222,7 @@ public class CrawlerMain extends ProcessMainClass {
 
         logger.info("Loaded {} domains", crawlSpecRecords.size());
 
-        // Shuffle the domains to ensure we get a good mix of domains in each crawl,
-        // so that e.g. the big domains don't get all crawled at once, or we end up
-        // crawling the same server in parallel from different subdomains...
-        Collections.shuffle(crawlSpecRecords);
+        crawlSpecRecords.sort(crawlSpecArrangement(crawlSpecRecords));
 
         // First a validation run to ensure the file is all good to parse
         if (crawlSpecRecords.isEmpty()) {
@@ -248,44 +246,35 @@ public class CrawlerMain extends ProcessMainClass {
             // List of deferred tasks used to ensure beneficial scheduling of domains with regard to DomainLocks,
             // merely shuffling the domains tends to lead to a lot of threads being blocked waiting for a semphore,
             // this will more aggressively attempt to schedule the jobs to avoid blocking
-            List<CrawlTask> deferredTasks = new LinkedList<>();
+            List<CrawlTask> taskList = new ArrayList<>();
 
-            // Create crawl tasks and submit them to the pool for execution
+            // Create crawl tasks
             for (CrawlSpecRecord crawlSpec : crawlSpecRecords) {
-                if (workLog.isJobFinished(crawlSpec.domain()))
+                if (workLog.isJobFinished(crawlSpec.domain))
                     continue;
 
-                // Add to the end of the deferral list
-                deferredTasks.addLast(new CrawlTask(
+                var task = new CrawlTask(
                         crawlSpec,
                         anchorTagsSource,
                         outputDir,
                         warcArchiver,
                         domainStateDb,
-                        workLog));
+                        workLog);
 
-                // Start every task we currently can from the deferral list
-                deferredTasks.removeIf(task -> {
-                    if (task.canRun()) {
-                        if (pendingCrawlTasks.putIfAbsent(crawlSpec.domain(), task) != null) {
-                            return true; // task has already run, duplicate in crawl specs
-                        }
-
-                        // This blocks the caller when the pool is full
-                        pool.submitQuietly(task);
-                        return true;
-                    }
-
-                    return false;
-                });
+                // Try to run immediately, to avoid unnecessarily keeping the entire work set in RAM
+                if (!trySubmitDeferredTask(task)) {
+                    // Otherwise add to the taskList for deferred execution
+                    taskList.add(task);
+                }
             }
 
-            // Schedule any lingering tasks for immediate execution
-            for (var task : deferredTasks) {
-                if (pendingCrawlTasks.putIfAbsent(task.domain, task) != null)
-                    continue;
+             // Schedule viable tasks for execution until list is empty
+            while (!taskList.isEmpty()) {
+                taskList.removeIf(this::trySubmitDeferredTask);
 
-                pool.submitQuietly(task);
+                // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                // we might have no new viable tasks to run for hours on end
+                TimeUnit.MILLISECONDS.sleep(50);
             }
 
             logger.info("Shutting down the pool, waiting for tasks to complete...");
@@ -312,6 +301,50 @@ public class CrawlerMain extends ProcessMainClass {
         }
     }
 
+    /** Create a comparator that sorts the crawl specs in a way that is beneficial for the crawl,
+     * we want to enqueue domains that tend ro be large and have common top domains first,
+     * but otherwise have a random order.
+     * <p></p>
+     * Note, we can't use hash codes for randomization as it is not desirable to have the same order
+     * every time the process is restarted (and CrawlSpecRecord is a record, which defines equals and
+     * hashcode based on the fields).
+     * */
+    private Comparator<CrawlSpecRecord> crawlSpecArrangement(List<CrawlSpecRecord> records) {
+        Random r = new Random();
+        Map<String, Integer> randomOrder = new HashMap<>(records.size());
+
+        for (var spec : records) {
+            randomOrder.put(spec.domain, r.nextInt());
+        }
+
+        return Comparator.comparing((CrawlSpecRecord spec) -> spec.domain.contains(".edu"))
+                .reversed()
+                .thenComparing(spec -> randomOrder.get(spec.domain))
+                .thenComparing(Record::hashCode); // non-deterministic tie-breaker to
+    }
+
+    /** Submit a task for execution if it can be run, returns true if it was submitted
+     * or if it can be discarded */
+    private boolean trySubmitDeferredTask(CrawlTask task) {
+        if (!task.canRun()) {
+            return false;
+        }
+
+        if (pendingCrawlTasks.putIfAbsent(task.domain, task) != null) {
+            return true; // task has already run, duplicate in crawl specs
+        }
+
+        try {
+            // This blocks the caller when the pool is full
+            pool.submitQuietly(task);
+            return true;
+        }
+        catch (RuntimeException ex) {
+            logger.error("Failed to submit task " + task.domain, ex);
+            return false;
+        }
+    }
+
     public void runForSingleDomain(String targetDomainName, FileStorageId fileStorageId) throws Exception {
         runForSingleDomain(targetDomainName, fileStorageService.getStorage(fileStorageId).asPath());
     }
@@ -377,6 +410,11 @@ public class CrawlerMain extends ProcessMainClass {
         @Override
         public void run() throws Exception {
 
+            if (workLog.isJobFinished(domain)) { // No-Op
+                logger.info("Omitting task {}, as it is already run", domain);
+                return;
+            }
+
             Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
             Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
             Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
@@ -431,7 +469,7 @@ public class CrawlerMain extends ProcessMainClass {
                 logger.error("Error fetching domain " + domain, e);
             }
             finally {
-                // We don't need to double-count these; it's also kept int he workLog
+                // We don't need to double-count these; it's also kept in the workLog
                 pendingCrawlTasks.remove(domain);
                 Thread.currentThread().setName("[idle]");
 
@@ -522,7 +560,7 @@ public class CrawlerMain extends ProcessMainClass {
     //
     // This must be synchronized as chewing through parquet files in parallel leads to enormous memory overhead
     private synchronized Path migrateParquetData(Path inputPath, String domain, Path crawlDataRoot) throws IOException {
-        if (!inputPath.endsWith(".parquet")) {
+        if (!inputPath.toString().endsWith(".parquet")) {
             return inputPath;
         }
 

code/processes/crawling-process/java/nu/marginalia/crawl/DomainStateDb.java

@@ -1,5 +1,8 @@
 package nu.marginalia.crawl;
 
+import com.google.inject.Inject;
+import nu.marginalia.storage.FileStorageService;
+import nu.marginalia.storage.model.FileStorageType;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -63,7 +66,29 @@ public class DomainStateDb implements AutoCloseable {
 
     public record FaviconRecord(String contentType, byte[] imageData) {}
 
-    public DomainStateDb(Path filename) throws SQLException {
+    @Inject
+    public DomainStateDb(FileStorageService fileStorageService) throws SQLException {
+        this(findFilename(fileStorageService));
+    }
+
+    private static Path findFilename(FileStorageService fileStorageService) throws SQLException {
+        var fsId = fileStorageService.getOnlyActiveFileStorage(FileStorageType.CRAWL_DATA);
+
+        if (fsId.isPresent()) {
+            var fs = fileStorageService.getStorage(fsId.get());
+            return fs.asPath().resolve("domainstate.db");
+        }
+        else {
+            return null;
+        }
+    }
+
+    public DomainStateDb(@Nullable Path filename) throws SQLException {
+        if (null == filename) {
+            connection = null;
+            return;
+        }
+
         String sqliteDbString = "jdbc:sqlite:" + filename.toString();
         connection = DriverManager.getConnection(sqliteDbString);
 
@@ -90,11 +115,18 @@ public class DomainStateDb implements AutoCloseable {
 
     @Override
     public void close() throws SQLException {
-        connection.close();
+        if (connection != null) {
+            connection.close();
+        }
     }
 
+    public boolean isAvailable() {
+        return connection != null;
+    }
 
     public void saveIcon(String domain, FaviconRecord faviconRecord) {
+        if (connection == null) throw new IllegalStateException("No connection to domainstate db");
+
         try (var stmt = connection.prepareStatement("""
                 INSERT OR REPLACE INTO favicon (domain, contentType, icon)
                        VALUES(?, ?, ?)
@@ -110,6 +142,9 @@ public class DomainStateDb implements AutoCloseable {
     }
 
     public Optional<FaviconRecord> getIcon(String domain) {
+        if (connection == null)
+            return Optional.empty();
+
         try (var stmt = connection.prepareStatement("SELECT contentType, icon FROM favicon WHERE DOMAIN = ?")) {
             stmt.setString(1, domain);
             var rs = stmt.executeQuery();
@@ -130,6 +165,8 @@ public class DomainStateDb implements AutoCloseable {
     }
 
     public void save(SummaryRecord record) {
+        if (connection == null) throw new IllegalStateException("No connection to domainstate db");
+
         try (var stmt = connection.prepareStatement("""
                 INSERT OR REPLACE INTO summary (domain, lastUpdatedEpochMs, state, stateDesc, feedUrl)
                 VALUES (?, ?, ?, ?, ?)
@@ -146,6 +183,9 @@ public class DomainStateDb implements AutoCloseable {
     }
 
     public Optional<SummaryRecord> get(String domainName) {
+        if (connection == null)
+            return Optional.empty();
+
         try (var stmt = connection.prepareStatement("""
                 SELECT domain, lastUpdatedEpochMs, state, stateDesc, feedUrl
                 FROM summary

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/HttpFetcherImpl.java

@@ -30,6 +30,7 @@ import java.net.http.HttpTimeoutException;
 import java.time.Duration;
 import java.util.*;
 import java.util.concurrent.Executors;
+import java.util.concurrent.Semaphore;
 import java.util.zip.GZIPInputStream;
 
 
@@ -60,7 +61,7 @@ public class HttpFetcherImpl implements HttpFetcher {
                 .cookieHandler(cookies)
                 .followRedirects(HttpClient.Redirect.NORMAL)
                 .connectTimeout(Duration.ofSeconds(8))
-                .executor(Executors.newCachedThreadPool())
+                .executor(Executors.newVirtualThreadPerTaskExecutor())
                 .build();
     }
 
@@ -116,7 +117,7 @@ public class HttpFetcherImpl implements HttpFetcher {
 
         for (int tries = 0;; tries++) {
             try {
-                var rsp = client.send(head, HttpResponse.BodyHandlers.discarding());
+                var rsp = SendLock.wrapSend(client, head, HttpResponse.BodyHandlers.discarding());
                 EdgeUrl rspUri = new EdgeUrl(rsp.uri());
 
                 if (!Objects.equals(rspUri.domain, url.domain)) {
@@ -153,7 +154,7 @@ public class HttpFetcherImpl implements HttpFetcher {
                     .timeout(requestTimeout)
                     ;
 
-                var rsp = client.send(headBuilder.build(), HttpResponse.BodyHandlers.discarding());
+                var rsp = SendLock.wrapSend(client, headBuilder.build(), HttpResponse.BodyHandlers.discarding());
                 var headers = rsp.headers();
 
                 var contentTypeHeader = headers.firstValue("Content-Type").orElse(null);
@@ -229,21 +230,24 @@ public class HttpFetcherImpl implements HttpFetcher {
 
         contentTags.paint(getBuilder);
 
-        HttpFetchResult result = warcRecorder.fetch(client, getBuilder.build());
+        try (var sl = new SendLock()) {
+            HttpFetchResult result = warcRecorder.fetch(client, getBuilder.build());
 
-        if (result instanceof HttpFetchResult.ResultOk ok) {
-            if (ok.statusCode() == 429) {
-                throw new RateLimitException(Objects.requireNonNullElse(ok.header("Retry-After"), "1"));
-            }
-            if (ok.statusCode() == 304) {
-                return new HttpFetchResult.Result304Raw();
-            }
-            if (ok.statusCode() == 200) {
-                return ok;
+            if (result instanceof HttpFetchResult.ResultOk ok) {
+                if (ok.statusCode() == 429) {
+                    throw new RateLimitException(Objects.requireNonNullElse(ok.header("Retry-After"), "1"));
+                }
+                if (ok.statusCode() == 304) {
+                    return new HttpFetchResult.Result304Raw();
+                }
+                if (ok.statusCode() == 200) {
+                    return ok;
+                }
             }
+
+            return result;
         }
 
-        return result;
     }
 
     @Override
@@ -317,22 +321,28 @@ public class HttpFetcherImpl implements HttpFetcher {
                 .timeout(requestTimeout)
                 .build();
 
-        var response = client.send(getRequest, HttpResponse.BodyHandlers.ofInputStream());
-        if (response.statusCode() != 200) {
-            return new SitemapResult.SitemapError();
-        }
-
-        try (InputStream inputStream = response.body()) {
-
-            InputStream parserStream;
-            if (sitemapUrl.path.endsWith(".gz")) {
-                parserStream = new GZIPInputStream(inputStream);
-            }
-            else {
-                parserStream = inputStream;
+        try (var sl = new SendLock()) {
+            var response = client.send(getRequest, HttpResponse.BodyHandlers.ofInputStream());
+            if (response.statusCode() != 200) {
+                return new SitemapResult.SitemapError();
+            }
+
+            Document parsedSitemap;
+
+            try (InputStream inputStream = response.body()) {
+                InputStream parserStream;
+                if (sitemapUrl.path.endsWith(".gz")) {
+                    parserStream = new GZIPInputStream(inputStream);
+                } else {
+                    parserStream = inputStream;
+                }
+
+                parsedSitemap = Jsoup.parse(parserStream, "UTF-8", sitemapUrl.toString(), Parser.xmlParser());
+            }
+            finally {
+                sl.close();
             }
 
-            Document parsedSitemap = Jsoup.parse(parserStream, "UTF-8", sitemapUrl.toString(), Parser.xmlParser());
             if (parsedSitemap.childrenSize() == 0) {
                 return new SitemapResult.SitemapError();
             }
@@ -386,7 +396,7 @@ public class HttpFetcherImpl implements HttpFetcher {
     }
 
     private Optional<SimpleRobotRules> fetchAndParseRobotsTxt(EdgeUrl url, WarcRecorder recorder) {
-        try {
+        try (var sl = new SendLock()) {
             var getRequest = HttpRequest.newBuilder()
                     .GET()
                     .uri(url.asURI())
@@ -429,5 +439,30 @@ public class HttpFetcherImpl implements HttpFetcher {
             }
         }
     }
+
+}
+
+class SendLock implements AutoCloseable {
+
+    private static final Semaphore maxConcurrentRequests = new Semaphore(Integer.getInteger("crawler.maxConcurrentRequests", 512));
+    boolean closed = false;
+
+    public SendLock() {
+        maxConcurrentRequests.acquireUninterruptibly();
+    }
+
+    public static <T> HttpResponse<T> wrapSend(HttpClient client, HttpRequest request, HttpResponse.BodyHandler<T> handler) throws IOException, InterruptedException {
+        try (var lock = new SendLock()) {
+            return client.send(request, handler);
+        }
+    }
+
+    @Override
+    public void close() {
+        if (!closed) {
+            maxConcurrentRequests.release();
+            closed = true;
+        }
+    }
 }
 

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/CrawlDelayTimer.java

@@ -3,6 +3,7 @@ package nu.marginalia.crawl.retreival;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 
 import java.time.Duration;
+import java.util.concurrent.ThreadLocalRandom;
 
 import static java.lang.Math.max;
 import static java.lang.Math.min;
@@ -53,12 +54,13 @@ public class CrawlDelayTimer {
     public void waitFetchDelay(long spentTime) {
         long sleepTime = delayTime;
 
+        long jitter = ThreadLocalRandom.current().nextLong(0, 150);
         try {
             if (sleepTime >= 1) {
                 if (spentTime > sleepTime)
                     return;
 
-                Thread.sleep(min(sleepTime - spentTime, 5000));
+                Thread.sleep(min(sleepTime - spentTime, 5000) + jitter);
             } else {
                 // When no crawl delay is specified, lean toward twice the fetch+process time,
                 // within sane limits. This means slower servers get slower crawling, and faster
@@ -71,17 +73,17 @@ public class CrawlDelayTimer {
                 if (spentTime > sleepTime)
                     return;
 
-                Thread.sleep(sleepTime - spentTime);
+                Thread.sleep(sleepTime - spentTime + jitter);
             }
 
             if (slowDown) {
                 // Additional delay when the server is signalling it wants slower requests
-                Thread.sleep(DEFAULT_CRAWL_DELAY_MIN_MS);
+                Thread.sleep(DEFAULT_CRAWL_DELAY_MIN_MS + jitter);
             }
         }
         catch (InterruptedException e) {
             Thread.currentThread().interrupt();
-            throw new RuntimeException();
+            throw new RuntimeException("Interrupted", e);
         }
     }
 }

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/CrawlerRetreiver.java

@@ -108,6 +108,11 @@ public class CrawlerRetreiver implements AutoCloseable {
                     DomainStateDb.SummaryRecord summaryRecord = sniffRootDocument(probedUrl, delayTimer);
                     domainStateDb.save(summaryRecord);
 
+                    if (Thread.interrupted()) {
+                        // There's a small chance we're interrupted during the sniffing portion
+                        throw new InterruptedException();
+                    }
+
                     // Play back the old crawl data (if present) and fetch the documents comparing etags and last-modified
                     if (crawlerRevisitor.recrawl(oldCrawlData, robotsRules, delayTimer) > 0) {
                         // If we have reference data, we will always grow the crawl depth a bit
@@ -140,7 +145,6 @@ public class CrawlerRetreiver implements AutoCloseable {
                             CrawlDelayTimer delayTimer,
                             DomainLinks domainLinks) {
 
-
         // Add external links to the crawl frontier
         crawlFrontier.addAllToQueue(domainLinks.getUrls(rootUrl.proto));
 
@@ -289,6 +293,10 @@ public class CrawlerRetreiver implements AutoCloseable {
         }
         catch (Exception ex) {
             logger.error("Error configuring link filter", ex);
+            if (Thread.interrupted()) {
+                Thread.currentThread().interrupt();
+                return DomainStateDb.SummaryRecord.forError(domain, "Crawler Interrupted", ex.getMessage());
+            }
         }
         finally {
             crawlFrontier.addVisited(rootUrl);

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/revisit/CrawlerRevisitor.java

@@ -46,6 +46,10 @@ public class CrawlerRevisitor {
                 break;
             }
 
+            if (Thread.interrupted()) {
+                throw new InterruptedException();
+            }
+
             var urlMaybe = EdgeUrl.parse(doc.url);
             if (urlMaybe.isEmpty())
                 continue;

code/processes/crawling-process/model/java/nu/marginalia/io/SerializableCrawlDataStream.java

@@ -42,18 +42,20 @@ public interface SerializableCrawlDataStream extends AutoCloseable {
     {
 
         String fileName = fullPath.getFileName().toString();
-        if (fileName.endsWith(".parquet")) {
+
+        if (fileName.endsWith(".slop.zip")) {
             try {
-                return new ParquetSerializableCrawlDataStream(fullPath);
+                return new SlopSerializableCrawlDataStream(fullPath);
             } catch (Exception ex) {
                 logger.error("Error reading domain data from " + fullPath, ex);
                 return SerializableCrawlDataStream.empty();
             }
         }
 
-        if (fileName.endsWith(".slop.zip")) {
+        else if (fileName.endsWith(".parquet")) {
+            logger.error("Opening deprecated parquet-style crawl data stream", new Exception());
             try {
-                return new SlopSerializableCrawlDataStream(fullPath);
+                return new ParquetSerializableCrawlDataStream(fullPath);
             } catch (Exception ex) {
                 logger.error("Error reading domain data from " + fullPath, ex);
                 return SerializableCrawlDataStream.empty();

code/services-application/search-service/build.gradle

@@ -41,6 +41,7 @@ dependencies {
 
     implementation project(':code:functions:live-capture:api')
     implementation project(':code:functions:math:api')
+    implementation project(':code:functions:favicon:api')
     implementation project(':code:functions:domain-info:api')
     implementation project(':code:functions:search-query:api')
 

code/services-application/search-service/java/nu/marginalia/search/SearchService.java

@@ -3,10 +3,14 @@ package nu.marginalia.search;
 import com.google.inject.Inject;
 import io.jooby.Context;
 import io.jooby.Jooby;
+import io.jooby.MediaType;
 import io.jooby.StatusCode;
 import io.prometheus.client.Counter;
 import io.prometheus.client.Histogram;
 import nu.marginalia.WebsiteUrl;
+import nu.marginalia.api.favicon.FaviconClient;
+import nu.marginalia.db.DbDomainQueries;
+import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.search.svc.*;
 import nu.marginalia.service.discovery.property.ServicePartition;
 import nu.marginalia.service.server.BaseServiceParams;
@@ -15,11 +19,14 @@ import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import java.util.List;
+import java.util.NoSuchElementException;
 
 public class SearchService extends JoobyService {
 
     private final WebsiteUrl websiteUrl;
     private final SearchSiteSubscriptionService siteSubscriptionService;
+    private final FaviconClient faviconClient;
+    private final DbDomainQueries domainQueries;
 
     private static final Logger logger = LoggerFactory.getLogger(SearchService.class);
     private static final Histogram wmsa_search_service_request_time = Histogram.build()
@@ -43,6 +50,8 @@ public class SearchService extends JoobyService {
                          SearchSiteInfoService siteInfoService,
                          SearchCrosstalkService crosstalkService,
                          SearchBrowseService searchBrowseService,
+                         FaviconClient faviconClient,
+                         DbDomainQueries domainQueries,
                          SearchQueryService searchQueryService)
     throws Exception {
         super(params,
@@ -58,6 +67,8 @@ public class SearchService extends JoobyService {
         this.websiteUrl = websiteUrl;
 
         this.siteSubscriptionService = siteSubscriptionService;
+        this.faviconClient = faviconClient;
+        this.domainQueries = domainQueries;
     }
 
     @Override
@@ -71,6 +82,32 @@ public class SearchService extends JoobyService {
         jooby.get("/site/https://*", this::handleSiteUrlRedirect);
         jooby.get("/site/http://*", this::handleSiteUrlRedirect);
 
+        String emptySvg = "<svg xmlns=\"http://www.w3.org/2000/svg\"></svg>";
+        jooby.get("/site/{domain}/favicon", ctx -> {
+            String domain = ctx.path("domain").value();
+            logger.info("Finding icon for domain {}", domain);
+            try {
+                DbDomainQueries.DomainIdWithNode domainIdWithNode = domainQueries.getDomainIdWithNode(new EdgeDomain(domain));
+                var faviconMaybe = faviconClient.getFavicon(domain, domainIdWithNode.nodeAffinity());
+
+                if (faviconMaybe.isEmpty()) {
+                    ctx.setResponseType(MediaType.valueOf("image/svg+xml"));
+                    return emptySvg;
+                } else {
+                    var favicon = faviconMaybe.get();
+
+                    ctx.responseStream(MediaType.valueOf(favicon.contentType()), consumer -> {
+                        consumer.write(favicon.bytes());
+                    });
+                }
+            }
+            catch (NoSuchElementException ex) {
+                ctx.setResponseType(MediaType.valueOf("image/svg+xml"));
+                return emptySvg;
+            }
+            return "";
+        });
+
         jooby.before((Context ctx) -> {
             ctx.setAttribute(startTimeAttribute, System.nanoTime());
         });

code/services-application/search-service/java/nu/marginalia/search/model/SearchFilters.java

@@ -81,6 +81,7 @@ public class SearchFilters {
                             ),
                             List.of(
                                     new Filter("Vintage", "fa-clock-rotate-left", SearchProfile.VINTAGE, parameters),
+                                    new Filter("Small Web", "fa-minus", SearchProfile.SMALLWEB, parameters),
                                     new Filter("Plain Text", "fa-file", SearchProfile.PLAIN_TEXT, parameters),
                                     new Filter("Tilde", "fa-house", SearchProfile.TILDE, parameters)
                             ),

code/services-application/search-service/resources/jte/explore/main.jte

@@ -10,7 +10,7 @@
 
 @template.part.head(title = "Marginalia Search - Explore")
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans ">
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans ">
 
 @template.part.navbar(navbar = navbar)
 
@@ -23,7 +23,7 @@
 </header>
 
 <div class="max-w-[1400px] mx-auto flex flex-col gap-1 place-items-center">
-    <div class="border dark:border-gray-600 bg-white dark:bg-gray-800 dark:text-gray-100 my-4 p-3 rounded overflow-hidden flex flex-col space-y-4">
+    <div class="border border-gray-300 dark:border-gray-600 bg-white dark:bg-gray-800 dark:text-gray-100 my-4 p-3 rounded overflow-hidden flex flex-col space-y-4">
         @if (results.hasFocusDomain())
             <div class="flex space-x-1">
                 <span>Showing websites similar to <a class="font-mono text-liteblue dark:text-blue-200" href="/site/${results.focusDomain()}"><i class="fas fa-globe"></i> <span class="underline">${results.focusDomain()}</span></a></span>
@@ -36,7 +36,7 @@
     </div>
     <div class="grid-cols-1 gap-4 sm:grid sm:grid-cols-1 md:grid-cols-3 xl:grid-cols-4 mx-auto sm:p-4">
         @for (BrowseResult result : results.results())
-            <div class="bg-white border dark:border-gray-600 dark:bg-gray-800 rounded overflow-hidden">
+            <div class="bg-white border border-gray-300 dark:border-gray-600 dark:bg-gray-800 rounded overflow-hidden">
                 <div class="bg-margeblue text-white p-2 flex space-x-4 text-sm">
                     <span class="break-words">${result.displayDomain()}</span>
                     <div class="grow"></div>

code/services-application/search-service/resources/jte/part/testenvHead.jte

@@ -9,6 +9,15 @@
                     nicotine: '#f8f8ee',
                     margeblue: '#3e5f6f',
                     liteblue: '#0066cc',
+                    bgblue: '#e5e9eb',
+                },
+                screens: {
+                    'coarsepointer': {
+                        'raw': '(pointer: coarse)'
+                    },
+                    'finepointer': {
+                        'raw': '(pointer: fine)'
+                    },
                 }
             },
             screens: {

code/services-application/search-service/resources/jte/serp/dict-lookup.jte

@@ -15,7 +15,7 @@
 
 @template.part.head(title = "Marginalia Search - " +  parameters.query())
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 @template.part.navbar(navbar = navbar)
 
 <div>

code/services-application/search-service/resources/jte/serp/error.jte

@@ -9,7 +9,7 @@
 
 @template.part.head(title = "Marginalia Search - Error")
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 
 @template.part.navbar(navbar = navbar)
 

code/services-application/search-service/resources/jte/serp/main.jte

@@ -11,7 +11,7 @@
 
 @template.part.head(title = "Marginalia Search - " +  results.getQuery())
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 @template.part.navbar(navbar = navbar)
 
 <div>
@@ -23,7 +23,7 @@
                 @template.serp.part.searchform(query = results.getParams().query(), profile = results.getProfile(), filters = results.getFilters())
             </div>
             <div class="grow"></div>
-            <button class="fixed bottom-10 right-5 sm:hidden text-sm bg-margeblue text-white p-4 rounded-xl active:text-slate-200" id="filter-button">
+            <button class="fixed bottom-10 right-5 finepointer:hidden md:hidden text-sm bg-margeblue text-white p-4 rounded-xl active:text-slate-200" id="filter-button">
                 <i class="fas fa-filter mr-3"></i>
                 Filters
             </button>

code/services-application/search-service/resources/jte/serp/part/footerHowto.jte

@@ -26,15 +26,15 @@
             It operates a bit like a clock, starting at the top and working its way around clockwise.</p>
 
         <div class="flex gap-4 place-items-middle">
-            @template.serp.part.matchogram(mask = 90)
+            @template.serp.part.matchogram(mask = 90, domain = "example.com")
             <div>This is by the beginning</div>
         </div>
         <div class="flex gap-4 place-items-middle">
-            @template.serp.part.matchogram(mask = 90L<<26)
+            @template.serp.part.matchogram(mask = 90L<<26, domain = "example.com")
             <div>This is in the middle</div>
         </div>
         <div class="flex gap-4 place-items-middle">
-            @template.serp.part.matchogram(mask = 5L<<48)
+            @template.serp.part.matchogram(mask = 5L<<48, domain = "example.com")
             <div>This is toward the end</div>
         </div>
 

code/services-application/search-service/resources/jte/serp/part/matchogram.jte

@@ -1,11 +1,13 @@
 @import java.util.stream.IntStream
 
 @param long mask
+@param String domain
 
-<svg width="40" height="40">
+<svg width="40" height="40"
+     style="background-image: url('/site/${domain}/favicon'); background-repeat: no-repeat; background-size: 16px 16px; background-position: center; ">
     <circle
-            cx="18"
-            cy="18"
+            cx="20"
+            cy="20"
             r="16"
             fill="none"
             stroke="#eee"
@@ -13,10 +15,10 @@
     />
 @for (int bit : IntStream.range(0, 56).filter(bit -> (mask & (1L << bit)) != 0).toArray())
             <line
-                    x1="${18 + 15*Math.sin(2 * Math.PI * bit / 56.)}"
-                    y1="${18 - 15*Math.cos(2 * Math.PI * bit / 56.)}"
-                    x2="${18 + 17*Math.sin(2 * Math.PI * bit / 56.)}"
-                    y2="${18 - 17*Math.cos(2 * Math.PI * bit / 56.)}"
+                    x1="${20 + 15*Math.sin(2 * Math.PI * bit / 56.)}"
+                    y1="${20 - 15*Math.cos(2 * Math.PI * bit / 56.)}"
+                    x2="${20 + 17*Math.sin(2 * Math.PI * bit / 56.)}"
+                    y2="${20 - 17*Math.cos(2 * Math.PI * bit / 56.)}"
                     stroke="#444"
                     stroke-width="2"
             />

code/services-application/search-service/resources/jte/serp/part/result.jte

@@ -12,7 +12,7 @@
             <div class="flex flex-col grow" >
                 <div class="flex flex-row space-x-2 place-items-center">
                     <div class="flex-0" title="Match density">
-                        @template.serp.part.matchogram(mask = result.first.positionsMask)
+                        @template.serp.part.matchogram(mask = result.first.positionsMask, domain=result.getFirst().url.domain.toString())
                     </div>
                     <div class="flex grow justify-between items-start">
                         <div class="flex-1">

code/services-application/search-service/resources/jte/serp/part/sidebar.jte

@@ -3,7 +3,7 @@
 
 @param SearchFilters filters
 
-<aside class="md:w-64 py-4 shrink-0 hidden sm:block">
+<aside class="md:w-64 py-4 shrink-0 hidden md:block finepointer:block">
     <div class="space-y-6 sticky top-4">
         <div class="bg-white dark:bg-gray-800 p-4 border dark:border-gray-600 border-gray-300">
             <h2 class="font-medium mb-3 flex items-center font-serif hidden md:block">
@@ -13,7 +13,7 @@
                 @for (List<SearchFilters.Filter> filterGroup : filters.getFilterGroups())
                     @for (SearchFilters.Filter filter : filterGroup)
                         <label class="flex items-center">
-                            <button title="${filter.displayName}" onclick="document.location='$unsafe{filter.url}'" class="flex-1 py-2 pl-2 rounded flex space-x-2 dark:has-[:checked]:bg-gray-950 has-[:checked]:bg-gray-100 has-[:checked]:text-slate-900 dark:has-[:checked]:text-slate-100 hover:bg-gray-50 dark:hover:bg-gray-950 bg-white dark:bg-gray-900 dark:border dark:border-gray-600 text-margeblue dark:text-slate-200 outline-1 active:outline">
+                            <button title="${filter.displayName}" onclick="document.location='$unsafe{filter.url}'" class="flex-1 py-2 pl-2 rounded flex space-x-2 dark:has-[:checked]:bg-gray-950 has-[:checked]:bg-gray-300 has-[:checked]:text-slate-900 dark:has-[:checked]:text-slate-100 hover:bg-gray-50 dark:hover:bg-gray-950 bg-white dark:bg-gray-900 dark:border dark:border-gray-600 text-margeblue dark:text-slate-200 outline-1 active:outline">
                                 @if (filter.current)
                                     <input type="checkbox" checked class="sr-only" aria-checked="true" />
                                 @else
@@ -38,7 +38,7 @@
             <div class="space-y-2">
                 @for (SearchFilters.SearchOption option : filters.searchOptions())
                     <label class="flex items-center">
-                            <button title="${option.name()}" onclick="document.location='$unsafe{option.getUrl()}'" class="flex-1 py-2 pl-2 rounded flex space-x-2 dark:has-[:checked]:bg-gray-950 has-[:checked]:bg-gray-100 has-[:checked]:text-slate-900 dark:has-[:checked]:text-slate-100 hover:bg-gray-50 dark:hover:bg-gray-950 bg-white dark:bg-gray-900 dark:border dark:border-gray-600 text-margeblue dark:text-slate-200 outline-1 active:outline">
+                            <button title="${option.name()}" onclick="document.location='$unsafe{option.getUrl()}'" class="flex-1 py-2 pl-2 rounded flex space-x-2 dark:has-[:checked]:bg-gray-950 has-[:checked]:bg-gray-300 has-[:checked]:text-slate-900 dark:has-[:checked]:text-slate-100 hover:bg-gray-50 dark:hover:bg-gray-950 bg-white dark:bg-gray-900 dark:border dark:border-gray-600 text-margeblue dark:text-slate-200 outline-1 active:outline">
                                 @if (option.isSet())
                                     <input type="checkbox" checked class="sr-only" aria-checked="true" />
                                 @else

code/services-application/search-service/resources/jte/serp/start.jte

@@ -15,7 +15,7 @@
 
 @template.part.head(title = "Marginalia Search", allowIndexing = true)
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 
 @template.part.navbar(navbar = navbar)
 
@@ -32,18 +32,14 @@
 
 @if (model.news().isEmpty())
 <div class="max-w-7xl mx-auto flex flex-col space-y-4 fill-w">
-    <div class="border dark:border-gray-600 dark:bg-gray-800 bg-white rounded p-2 m-4 ">
+    <div class="border border-gray-300 border-gray-100 dark:border-gray-600 dark:bg-gray-800 bg-white rounded p-2 m-4 ">
         <div class="text-slate-700 dark:text-white text-sm p-4">
-            <div class="fas fa-gift mr-1 text-margeblue dark:text-slate-200"></div>
-            This is the new design and home of Marginalia Search.
-            You can read about what this entails <a href="https://about.marginalia-search.com/article/redesign/"  class="underline text-liteblue dark:text-blue-200">here</a>.
-        <p class="my-4"></p>
-        The old version of Marginalia Search remains available at
-        <a href="https://old-search.marginalia.nu/" class="underline text-liteblue dark:text-blue-200">https://old-search.marginalia.nu/</a>.
+        The old version of Marginalia Search remains available
+        <a href="https://old-search.marginalia.nu/" class="underline text-liteblue dark:text-blue-200">here</a>.
         </div>
     </div>
     <div class="mx-auto flex flex-col sm:flex-row my-4 sm:space-x-2 space-y-2 sm:space-y-0 w-full md:w-auto px-2">
-    <div class="flex flex-col border dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6 space-y-3">
+    <div class="flex flex-col border border-gray-300 dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6 space-y-3">
         <div><i class="fas fa-sailboat mx-2 text-margeblue dark:text-slate-200"></i>Explore the Web</div>
         <ul class="list-disc ml-6 text-slate-700 dark:text-white text-xs leading-5">
             <li>Prioritizes non-commercial content</li>
@@ -52,7 +48,7 @@
         </ul>
     </div>
 
-    <div class="flex flex-col border dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6  space-y-3 ">
+    <div class="flex flex-col border border-gray-300 dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6  space-y-3 ">
         <div><i class="fas fa-hand-holding-hand mx-2 text-margeblue dark:text-slate-200"></i>Open Source</div>
         <ul class="list-disc ml-6 text-slate-700 dark:text-white text-xs leading-5">
            <li>Custom index and crawler software</li>
@@ -65,7 +61,7 @@
         </div>
     </div>
 
-    <div class="flex flex-col border dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6  space-y-3 ">
+    <div class="flex flex-col border border-gray-300 dark:border-gray-600 rounded overflow-hidden dark:bg-gray-800 bg-white p-6  space-y-3 ">
         <div><i class="fas fa-lock mx-2 text-margeblue dark:text-slate-200"></i> Privacy by default</div>
         <ul class="list-disc ml-6 text-slate-700 dark:text-white text-xs leading-5">
             <li>Filter out tracking and adtech</li>

code/services-application/search-service/resources/jte/serp/unit-conversion.jte

@@ -13,7 +13,7 @@
 
 @template.part.head(title = "Marginalia Search - " +  parameters.query())
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 @template.part.navbar(navbar = navbar)
 
 <div>

code/services-application/search-service/resources/jte/siteinfo/crosstalk.jte

@@ -11,7 +11,7 @@
 
 @template.part.head(title = "Marginalia Search - " + model.domainA() + "/" + model.domainB())
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 
 @template.part.navbar(navbar = navbar)
 

code/services-application/search-service/resources/jte/siteinfo/main.jte

@@ -9,7 +9,7 @@
 
 @template.part.head(title = "Marginalia Search - " + model.domain())
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 
 @template.part.navbar(navbar = navbar)
 

code/services-application/search-service/resources/jte/siteinfo/part/linkedDomains.jte

@@ -7,7 +7,7 @@
 
 @if (!list.isEmpty())
 
-    <div class="bg-white dark:bg-gray-800 shadow-sm rounded overflow-hidden border dark:border-gray-600">
+    <div class="bg-white dark:bg-gray-800 shadow-sm rounded overflow-hidden border border-gray-300 dark:border-gray-600">
         <div class="px-4 py-2 bg-margeblue text-white border-b border-gray-200 dark:border-gray-600 flex place-items-baseline">
             <h2 class="text-md">${title}</h2>
             <div class="grow"></div>

code/services-application/search-service/resources/jte/siteinfo/start.jte

@@ -9,11 +9,11 @@
 
 @template.part.head(title = "Marginalia Search - Site Viewer")
 
-<body class="min-h-screen bg-slate-100 dark:bg-gray-900 dark:text-white font-sans " >
+<body class="min-h-screen bg-bgblue dark:bg-gray-900 dark:text-white font-sans " >
 
 @template.part.navbar(navbar = navbar)
 
-<header class="border-b border-gray-300 dark:border-gray-600 bg-white dark:bg-gray-800 shadow-md">
+<header class="border-b border-gray-300 border-gray-300 dark:border-gray-600 bg-white dark:bg-gray-800 shadow-md">
     <div class="max-w-[1400px] mx-auto px-4 py-4">
         <h1 class="text-base md:text-xl mr-2 md:mr-8 font-serif">View Site Information</h1>
     </div>
@@ -22,7 +22,7 @@
 <div class="max-w-[1000px] mx-auto flex gap-4 flex-col md:flex-row place-items-center md:place-items-start p-4">
 
 
-    <div class="border dark:border-gray-600 rounded md:my-4 overflow-hidden bg-white dark:bg-gray-800 flex flex-col space-y-2 flex-1">
+    <div class="border border-gray-300 dark:border-gray-600 rounded md:my-4 overflow-hidden bg-white dark:bg-gray-800 flex flex-col space-y-2 flex-1">
         <div class="bg-margeblue text-white p-2 text-sm mb-2">View Site Information</div>
 
         <p class="mx-4">This utility lets you explore what the search engine knows about the web,
@@ -45,7 +45,7 @@
     </div>
 
     @if (!model.domains().isEmpty())
-    <div class="border dark:border-gray-600 rounded md:my-4 overflow-hidden w-full md:w-auto">
+    <div class="border border-gray-300 dark:border-gray-600 rounded md:my-4 overflow-hidden w-full md:w-auto">
         <div class="bg-margeblue text-white p-2 text-sm">Recently Discovered Domains</div>
 
 

code/services-application/search-service/resources/jte/siteinfo/view/backlinks.jte

@@ -8,17 +8,17 @@
 <div class="flex flex-col space-y-4 my-4  w-full">
 
 @if (backlinks.results().isEmpty())
-<div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm ">
+<div class="border border-gray-300 dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm ">
     The search engine isn't aware of any backlinks to ${backlinks.domain()}!
 </div>
 @else
-<div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
+<div class="border border-gray-300 dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
     Showing documents linking to ${backlinks.domain()}
 </div>
 @endif
 
 @for (GroupedUrlDetails group : backlinks.results())
-<div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden mx-4">
+<div class="border dark:border-gray-600 border-gray-300 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden mx-4">
     <div class="flex space-x-2 flex-row place-items-baseline bg-margeblue text-white p-2  text-md">
         <span class="fas fa-globe"></span>
         <a href="/site/${group.domain().toString()}">${group.domain().toString()}</a>

code/services-application/search-service/resources/jte/siteinfo/view/docs.jte

@@ -9,17 +9,17 @@
 <div class="flex flex-col space-y-4 my-4">
 
 @if (docs.results().isEmpty())
-    <div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
+    <div class="border border-gray-300  dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
         The search engine doesn't index any documents from ${docs.domain()}
     </div>
 @else
-    <div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
+    <div class="border border-gray-300 dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden p-4 mx-4 text-gray-800 text-sm">
         Showing documents from ${docs.domain()}
     </div>
 @endif
 
 @for (UrlDetails details : docs.results())
-<div class="border dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden mx-4">
+<div class="border border-gray-300 dark:border-gray-600 rounded bg-white dark:bg-gray-800 dark:text-white flex flex-col overflow-hidden mx-4">
     <div class="flex grow justify-between items-start p-4">
         <div class="flex-1">
             <h2 class="text-xl text-gray-800 dark:text-white font-serif mr-4">

code/services-application/search-service/resources/jte/siteinfo/view/overview.jte

@@ -8,9 +8,9 @@
 <!-- Main content -->
 
 <div class="flex-1 p-4 space-y-4 mx-auto w-full md:w-auto">
-    <div class="flex border dark:border-gray-600 rounded bg-white dark:bg-gray-800 flex-col space-y-4 pb-4 overflow-hidden md:max-w-lg" >
-        <div class="flex place-items-baseline space-x-2 p-2 text-md border-b dark:border-gray-600 bg-margeblue text-white">
-            <i class="fa fa-globe"></i>
+    <div class="flex border border-gray-300  dark:border-gray-600 rounded bg-white dark:bg-gray-800 flex-col space-y-4 pb-4 overflow-hidden md:max-w-lg" >
+        <div class="flex place-items-center space-x-2 p-2 text-md border-b dark:border-gray-600 bg-margeblue text-white">
+            <img src="/site/${siteInfo.domain()}/favicon" style="width: 16px; height: 16px; vertical-align: center">
             <span>${siteInfo.domain()}</span>
             <div class="grow">
             </div>

code/services-application/search-service/resources/jte/siteinfo/view/reportDomain.jte

@@ -4,7 +4,7 @@
 @param ReportDomain reportDomain
 
 <div class="flex-col mx-auto">
-<div class="max-w-2xl mx-auto bg-white dark:bg-gray-800 border dark:border-gray-600 rounded overflow-auto shadow-sm my-4 space-y-4 w-full">
+<div class="max-w-2xl mx-auto bg-white dark:bg-gray-800 border border-gray-300 dark:border-gray-600 rounded overflow-auto shadow-sm my-4 space-y-4 w-full">
     <div class="px-4 py-2 bg-margeblue text-white border-b border-gray-200 dark:border-gray-800">
         <h2 class="text-md">Report Domain Issue</h2>
     </div>

code/services-application/search-service/tailwind/tailwind.config.js

@@ -9,6 +9,15 @@ module.exports = {
         nicotine: '#f8f8ee',
         margeblue: '#3e5f6f',
         liteblue: '#0066cc',
+        bgblue: '#e5e9eb',
+      },
+      screens: {
+        'coarsepointer': {
+          'raw': '(pointer: coarse)'
+        },
+        'finepointer': {
+          'raw': '(pointer: fine)'
+        },
       }
     },
     screens: {

code/services-core/executor-service/build.gradle

@@ -42,6 +42,8 @@ dependencies {
     implementation project(':code:libraries:message-queue')
 
     implementation project(':code:functions:link-graph:api')
+    implementation project(':code:functions:favicon')
+    implementation project(':code:functions:favicon:api')
 
     implementation project(':code:processes:crawling-process:model')
     implementation project(':code:processes:crawling-process:model')

code/services-core/executor-service/java/nu/marginalia/executor/ExecutorSvc.java

@@ -2,6 +2,7 @@ package nu.marginalia.executor;
 
 import com.google.inject.Inject;
 import nu.marginalia.execution.*;
+import nu.marginalia.functions.favicon.FaviconGrpcService;
 import nu.marginalia.service.discovery.property.ServicePartition;
 import nu.marginalia.service.server.BaseServiceParams;
 import nu.marginalia.service.server.SparkService;
@@ -24,6 +25,7 @@ public class ExecutorSvc extends SparkService {
                        ExecutorCrawlGrpcService executorCrawlGrpcService,
                        ExecutorSideloadGrpcService executorSideloadGrpcService,
                        ExecutorExportGrpcService executorExportGrpcService,
+                       FaviconGrpcService faviconGrpcService,
                        ExecutionInit executionInit,
                        ExecutorFileTransferService fileTransferService) throws Exception {
         super(params,
@@ -31,7 +33,8 @@ public class ExecutorSvc extends SparkService {
                 List.of(executorGrpcService,
                         executorCrawlGrpcService,
                         executorSideloadGrpcService,
-                        executorExportGrpcService)
+                        executorExportGrpcService,
+                        faviconGrpcService)
             );
 
         this.executionInit = executionInit;

settings.gradle

@@ -16,7 +16,8 @@ include 'code:services-application:status-service'
 
 include 'code:functions:math'
 include 'code:functions:math:api'
-
+include 'code:functions:favicon'
+include 'code:functions:favicon:api'
 include 'code:functions:domain-info'
 include 'code:functions:domain-info:api'