Re: providing a libidn compatibility API

[Tim Ruehsen]

On Thursday, January 26, 2017 2:07:39 PM CET Nikos Mavrogiannopoulos wrote:
> I certainly do. Note that there few failures in the libidn testsuite,
> which I have put them in ifdef XXX. These are in tst_idna2.c, and
> tst_idna4.c. The tst_idna4.c failure has to do with ascii_to_8z in
> libidn2 allowing multiple dots, while the tst_idn2.c failures are
> beyond my knowledge.

As you say, libidn/idn does not allow empty labels, e.g. looking up 
"search..." would result in IDNA_INVALID_LENGTH.

Libidn2/idn2 currently does allow empty labels, there is no equivalent to 
IDNA_INVALID_LENGTH.

https://tools.ietf.org/html/rfc1034#section-3.5 defines a label having at least 
one character ("must start with a letter").
<label> ::= <letter> [ [ <ldh-str> ] <let-dig> ]

Some more constraints regarding A/U-labels are described at
https://tools.ietf.org/html/rfc5890#section-2.3.2

IMO, libidn2 should check these limits
We would need IDN2_LABEL_TOO_SHORT and a few tests.

Regards, Tim

signature.asc
Description: This is a digitally signed message part.