Re: [EAI] Proposed resolution of List-* and downgrade- Issues

["Frank Ellermann" <hmdmhdfmhdjmzdtjmzdtzktdkztdjz at gmail.com>]

Charles Lindsey wrote:

>> JFTR, s/as always with IRIs/as always with EAI/
[...]
> 1. You convert the KO18-R into unicode
> 2. You convert the unicode into UTF-8
> 3. You %-encode the UTF-8
> 4. Now you have the proper URI

> Of course if the context is headers within a UTF8SMTP
> message, then steps 1 and 2 are redundant.

Yes, that's what I meant.  And depending on the charset
steps 1 and 2 could be combined, KOI8-R is an example
where that works - it works for all SBCS with US-ASCII
as propoer subset.
  
>> you would use IDNA(bis) rules to convert this into
>> an URL:
>> Archived-At: <http://archive.xn--maana-pta.com/>

> RFC 3987 says you MAY (possibily even SHOULD) convert
> to punycode in this case, but you also MAY just %-encode
> it and hope whatever comes next knows how to punycode it
> before doing any DNS lookup.

The goal is to get a <reg-name> for DNS in many relevant
contexts:

| A registered name intended for lookup in the DNS uses
| the syntax defined in Section 3.5 of [RFC1034] and
| Section 2.1 of [RFC1123].

 [Digression, this remark in RFC 3986 kills hypothetical
  IDN TLDs again, until RFC 1123 section 2.1 is updated.]

| Such a name consists of a sequence of domain labels
| separated by ".",  each domain label starting and ending
| with an alphanumeric character and possibly also
| containing "-" characters.

Clearly no "percent" characters in a <reg-name> for DNS.

That is as it should be, DNS won't do the punycode magic,
legacy applications (i.e. almost all applications minus
some modern browsers, xmpp, or similar) also won't do it,
but RFC 3987 behind a barrage of MAY and SHOULD weasel
words "does" it.  Where it doesn't it cannot work, there
is no archive.ma%c3%B1ana.com in DNS.

Hopefully, nobody knows what this could be if it exists,
I'd bet on all possible malicious purposes.

But you are right - any serious 3987bis proposal has to
fix this strange "MAY or SHOULD" to a simple "MUST" wrt
DNS.  Nobody is going to register %-encoded raw UTF-8
CNAMEs for all IDN A-labels.  Only the opposition might
try that where it sees a potential profit.

> IRIs are not currently allowed in the 'http' scheme

Why do you think that http-IRIs are not allowed ?  My
assumption is that all URI-schemes can be also used as
IRI-schemes, in theory, down to 'mailto' as worst case.

But some document formats, notably HTML 4, XHTML 1, and
message/rfc822 header fields, allow only URIs.  And no
oddities with "percent" characters in a <reg-name> if
this is about the registry known as DNS.  

 Frank

_______________________________________________
IMA mailing list
IMA at ietf.org
https://www.ietf.org/mailman/listinfo/ima