Measure how often DNS hostnames aren't in preferred name form.

net/dns/dns_util.h

Index: net/dns/dns_util.h
diff --git a/net/dns/dns_util.h b/net/dns/dns_util.h
index 06cc3d6f8b731e5fe95f96b218afb339528d3036..8bde714f5c6fca9660556a20122946e511de43bf 100644
--- a/net/dns/dns_util.h
+++ b/net/dns/dns_util.h
@@ -27,6 +27,21 @@ NET_EXPORT_PRIVATE bool DNSDomainFromDot(const base::StringPiece& dotted,
 // Checks that a hostname is valid. Simple wrapper around DNSDomainFromDot.
 NET_EXPORT_PRIVATE bool IsValidDNSDomain(const base::StringPiece& dotted);
 
+// Returns true if the character is valid in a DNS hostname label, whether in
+// the first position or later in the label.
+//
+// This function asserts a looser form of the restrictions in RFC 1123 (section

[Ryan Sleevi, 2017/03/22 00:16:07]

https://tools.ietf.org/html/rfc7719#section-2 is meant to be the clarifying
document about whose rules matter :) And include the link back to 1123 (under
"Host name")

[palmer, 2017/03/22 01:03:33]

Done.

+// 2.1): hostnames can include characters a-z, A-Z, 0-9, -, and _, and any of
+// those characters (except -) are legal in the first position. The looser rules
+// are necessary to support service records (initial _), and non-compliant but
+// attested hostnames that include _. These looser rules also allow Punycode and
+// hence IDN.
+//
+// TODO(palmer): In the future, when we can remove support for invalid names,
+// this can be a private implementation detail of |DNSDomainFromDot|, and need
+// not be NET_EXPORT_PRIVATE.
+NET_EXPORT_PRIVATE bool IsValidLabelCharacter(char c, bool is_first_char);

[Ryan Sleevi, 2017/03/22 00:16:07]

pedantry: IsValidHostLabelCharacter? Since other labels can and do exceed the
host record (implied A/AAAA, such as SRVNames)

[palmer, 2017/03/22 01:03:34]

Done.

+
 // DNSDomainToString converts a domain in DNS format to a dotted string.
 // Excludes the dot at the end.
 NET_EXPORT_PRIVATE std::string DNSDomainToString(