@@ -220,0 +222,4 @@
+    xml_declaration: str | None,
+    rdf_root: str | None,
+    description_lines: list[str],
+) -> str:

@@ -220,0 +311,4 @@
+    depth = 0
+
+    for line in lines:
+        line_count += 1

@@ -220,0 +321,4 @@
+        # Extract XML declaration (line 1)
+        if stripped.startswith("<?xml"):
+            root_line = line
+            continue

@@ -220,0 +326,4 @@
+        # Extract root RDF element with namespaces (line 2)
+        if stripped.startswith("<rdf:RDF") and rdf_root_line is None:
+            rdf_root_line = line
+            continue

@@ -220,0 +334,4 @@
+            depth = 1
+            current_description = [line]
+            # Handle self-closing tags: <rdf:Description ... />
+            if stripped.rstrip().endswith("/>"):

@@ -220,0 +349,4 @@
+                if description_count >= chunk_size:
+                    yield triple_batch
+                    triple_batch = []
+                    description_count = 0

@@ -220,0 +358,4 @@
+
+            # Count opening/closing tags to track depth correctly
+            # Opening tags: <tag> or <tag ...> (not </tag or <tag/>)
+            # Find all <tag patterns (this excludes </tag> because regex requires [a-zA-Z_] after <)

@@ -220,0 +359,4 @@
+            # Count opening/closing tags to track depth correctly
+            # Opening tags: <tag> or <tag ...> (not </tag or <tag/>)
+            # Find all <tag patterns (this excludes </tag> because regex requires [a-zA-Z_] after <)
+            all_tag_starts = re.findall(r'<([a-zA-Z_][\w:.-]*)', stripped)

@@ -220,0 +364,4 @@
+            closing_tags = len(re.findall(r'</([a-zA-Z_][\w:.-]*)>', stripped))
+            # Self-closing tags: <tag ... /> (these don't change depth)
+            # Find tags that end with /> before the next >
+            self_closing_tags = len(re.findall(r'<([a-zA-Z_][\w:.-]*)[^>]*/>', stripped))

@@ -220,0 +367,4 @@
+            self_closing_tags = len(re.findall(r'<([a-zA-Z_][\w:.-]*)[^>]*/>', stripped))
+
+            # Opening tags = all tag starts - self-closing tags
+            # (all_tag_starts already excludes closing tags, so we only subtract self-closing)

@@ -220,0 +377,4 @@
+            # Check if Description element is closed
+            if "</rdf:Description>" in stripped:
+                # The closing tag was already accounted for in closing_tags above
+                if depth == 0:

@@ -220,0 +405,4 @@
+
+        if line_count % PROGRESS_LOG_INTERVAL == 0:
+            console.print(
+                f"[dim]Processed {line_count:,} lines, {description_count:,} descriptions...[/dim]"

@@ -227,3 +435,3 @@
     Args:
         lines: Iterator yielding text lines of RDF data
-        format: RDF serialization format (supported: nt, ntriples, turtle, ttl)
+        format: RDF serialization format (supported: nt, ntriples, turtle, ttl, xml, rdf, rdfxml)

@@ -252,3 +462,3 @@
 
 
-def _pick_zip_member(zf: zipfile.ZipFile, *, prefer_exts: tuple[str, ...] = (".ttl", ".nt")) -> str:
+def _pick_zip_member(zf: zipfile.ZipFile, *, prefer_exts: tuple[str, ...] = (".ttl", ".nt", ".rdf", ".xml")) -> str:

@@ -989,0 +1206,4 @@
+                "ntriples": "ntriples",
+            }
+            rdf_format = format_mapping.get(registry_format, registry_format)
+            console.print(f"[dim]Format auto-detected from registry: {registry_format} → {rdf_format}[/dim]")

@@ -1001,3 +1233,3 @@
             }
             rdf_format = format_map.get(ext, "turtle")
-            console.print(f"[dim]Format inferred from extension: {rdf_format}[/dim]")
+            console.print(f"[dim]Format inferred from extension: {ext} → {rdf_format}[/dim]")

@@ -27,0 +41,4 @@
+    if protocol == "ftp":
+        yield from stream_ftp_bytes(url, config=config)
+    elif protocol in ("http", "https"):
+        yield from stream_http_bytes_impl(url, headers=headers, start_byte=start_byte, config=config)

@@ -27,0 +43,4 @@
+    elif protocol in ("http", "https"):
+        yield from stream_http_bytes_impl(url, headers=headers, start_byte=start_byte, config=config)
+    else:
+        raise ValueError(f"Unsupported protocol: {protocol}. Supported: http, https, ftp")

@@ -76,0 +148,4 @@
+                raise ValueError(f"File not found on FTP server: {url}") from e
+            elif "timed out" in reason.lower():
+                # Retry on timeout
+                if attempt > cfg.max_retries:

@@ -76,0 +150,4 @@
+                # Retry on timeout
+                if attempt > cfg.max_retries:
+                    raise
+                sleep_s = min(cfg.backoff_base_s * (2 ** (attempt - 1)), cfg.backoff_max_s)

@@ -76,0 +156,4 @@
+                # Other errors: retry if we have attempts left
+                if attempt > cfg.max_retries:
+                    raise
+                sleep_s = min(cfg.backoff_base_s * (2 ** (attempt - 1)), cfg.backoff_max_s)

@@ -76,0 +159,4 @@
+                sleep_s = min(cfg.backoff_base_s * (2 ** (attempt - 1)), cfg.backoff_max_s)
+                time.sleep(sleep_s)
+
+        except Exception as e:

@@ -220,0 +302,4 @@
+        return triples
+    except Exception as e:
+        logger.error(f"Error parsing Description XML: {e}")
+        logger.debug(f"XML content (first 500 chars): {xml_content[:500]}")

@@ -21,1 +28,4 @@
+    return min(base_s * (2 ** (attempt - 1)), max_s)
+
+
 def stream_http_bytes(