Generating CIDs in Blink during MHTML serialization.

third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp

Index: third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp
diff --git a/third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp b/third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp
index cc67350bc8c7374af3a0676d2470fe7922747496..45846986e945786ee9c398bec4ab24c3b5d3be7f 100644
--- a/third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp
+++ b/third_party/WebKit/Source/platform/mhtml/MHTMLParser.cpp
@@ -400,4 +400,26 @@ ArchiveResource* MHTMLParser::subResourceAt(size_t index) const
     return m_resources[index].get();
 }
 
+// static
+KURL MHTMLParser::convertContentIDToURI(const String& contentID)
+{
+    // This function is based primarily on an example from rfc2557 in section
+    // 9.5, but also based on more normative parts of specs like:
+    // - rfc2557 - MHTML - section 8.3 - "Use of the Content-ID header and CID URLs"
+    // - rfc1738 - URL - section 4 (reserved scheme names;  includes "cid")
+    // - rfc2387 - multipart/related - section 3.4 - "Syntax" (cid := msg-id)
+    // - rfc0822 - msg-id = "<" addr-spec ">"; addr-spec = local-part "@" domain
+
+    if (contentID.length() <= 2)
+        return KURL();
+
+    if (!contentID.startsWith('<') || !contentID.endsWith('>'))
+        return KURL();
+
+    StringBuilder uriBuilder;
+    uriBuilder.append("cid:");
+    uriBuilder.append(contentID, 1, contentID.length() - 2);
+    return KURL(KURL(), uriBuilder.toString());

[dcheng, 2015/12/01 04:19:31]

Maybe use ParsedURLString? It's guaranteed that uriBuilder.toString() will be a
valid URL right?

[Łukasz Anforowicz, 2015/12/01 18:45:14]

I almost made the change you've suggested, but started having second thoughts
after realizing that I can't really add an assert that the url will be always
valid.  This is because |contentID| may come from an untrusted source (i.e. from
MHTML/MIME headers from an MHTML document).  Additionally, this doesn't feel
like a performance-sensitive code path (OTOH, this statement is based on an
armchair analysis rather than on data), so it might be better to just play it
safe.

Given above, I think it is better to leave it as is.  WDYT?

+}
+
 } // namespace blink