On Fri, 15 Aug 2003, Hollenbeck, Scott wrote: > > I guess the best way to summarize the behavior is by way of a > > pattern. I > > tested using values of the form: > > > > telnet://[a-z]{n}.com > > > > and found that, for values of n <= 62, the values was > > accepted, but when > > n >= 64, the value is rejected as being outside the value space for > > anyURI. It looks as if Xerces is imposing a length > > restriction on anyURI > > which has not been imposed in previous releases I have used > > (specifically, release 2.4.0 does not exhibit this behavior). > > > > I note that the values I am using, as can be seen from the pattern, > > contain only two components in the hostname, the first of > > which can be > > very long (and not necessarily very sensible :-) ). Here's an example: > > > > > > telnet://effortstandardsbesenseorganizationsecfordevelopedtoac > cessibleb.com > > I've checked the Schema REC and errata, as well as the relevant RFCs and > sections of XLink and XML Core, and can find no mention of a constraint > on the length of the anyURI datatype, on URIs in general, or on the > components of URIs. > > Have I missed something? Any feedback gratefully appreciated. > > RFC 1034 (sections 3.1 and 3.5) limits the length of domain name labels (the > strings between the dots) to a maximum of 63 characters. Perhaps that has > something to do with the behavior you're seeing.
This is precisely the behaviour being observed. RFC 2396, Section 3.2.2 Server-based Naming Authority states that "Hostnames take the form described in Section 3 of [RFC1034] and Section 2.1 of [RFC1123]: a sequence of domain labels separated by ".", each domain label starting and ending with an alphanumeric character and possibly also containing "-" characters. Hence the restriction on domain name label lengths. Among several malformed productions, the BNF for hostname is incorrect in RFC 2396. You can track the evolution of RFC 2396 in http://www.ietf.org/internet-drafts/draft-fielding-uri-rfc2396bis-03.txt. If it becomes an RFC, it will obsolete RFC 2396 and RFC 2732. The example URI given in the original post does not contain a valid hostname. The authority component does however match reg_name, however Registry-based Naming Authority has never been supported in the URI implementation. The example URI was passing as a false positive, and only affected edge cases such as the given example. I intend to add support for reg_name in the next release, though some future XML errata may remove it (The production is intended to be removed from URI for the reasons given in http://www.apache.org/~fielding/uri/rev-2002/issues.html#040-reg-name). > Scott Hollenbeck > VeriSign Naming and Directory Services > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > -- -------------------- Michael Glavassevich [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
