Stop me if I've misunderstood...

Paul Hoffman phoffman at imc.org
Fri Jul 10 20:22:17 CEST 2009


At 2:46 PM +0100 7/10/09, Gervase Markham wrote:
>On 09/07/09 15:47, Paul Hoffman wrote:
>>High-level: there is a difference between "misunderstood" and
>>"haven't read the WG archives nor the WG charter". It's not like we
>>haven't discussed each of these in detail before now.
>
>I'm sure you have. But, as I summarised for someone else late last night:
>
>From my point of view, in brief, here's how it looked: big phishing problem, group created to solve much of the issue at a protocol level, was looking good, banning lots of unnecessary characters etc., and also doing good stuff updating to the latest version of Unicode... and then loads of messages start flying backwards and forwards, the volume gets too much and when the dust settles, what seems to be important parts of IDNA2003 are on the floor in pieces, with people saying "but we _never_ even considered keeping _those_ bits... where have you been?"

- Phishing has never been a motivator for the WG

- You seem to think that "loads of messages" were recent. This topic has come up on the list in at least two batches in the past.

- Until about a month ago, the WG documents never had anything about IDNA2003-style mapping. So, this is not a case of them ending up on the floor, it's a case of us looking at them for the first time.

>>Please see the charter, then the (still-poorly-named) rationale
>>document.
>
>I agree with everything in the charter, except (it would now seem) the second half of the very last sentence in point iii).

Quite right. You may have missed the subtly of "It is not expected to continue to use Nameprep (RFC 3491).", but you cannot be blamed for that.

>Nothing there explains to me why removing the MUST status of normalizations and mappings is necessary, or even desirable.

...except that last sentence in point iii, which is explicit.

>There are lots of laudable goals there; some may claim that this removal is necessary to reach them, but the charter does not demonstrate this.

That is not the role of the charter.

>Having reviewed the rationale document, it appears that there may be some level of talking past each other here; if so, I apologise. It says:
>
>"Case-matching must be done, if desired, by IDN clients even though it wasn't done by ASCII- only DNS clients.  That situation was recognized in IDNA2003 and nothing in these specifications fundamentally changes it or could do so. In IDNA2003, all characters are case-folded and mapped by clients in a standardized step."
>
>Is this saying that case-folding and normalization remains (in overwhelming part) in IDNA2008 just as in IDNA2003?

No.

>If so, I am most encouraged. That was not the impression I got from recent mailing list traffic. But it does raise the question about what exactly is being removed, if not this? I am suddenly further confused.

See my earlier rants about the rationale document. Your confusion is justified.

>>Please see draft-ietf-idnabis-mappings-01.txt, which is what we are
>>currently discussing. It is the guidance that you seek.
>
>So what was a MUST part of the protocol is now just advisory?

There was never a MUST part of the IDNA2008 protocol. If you meant "...of the previous protocol", the answer is yes.

> Who, if anyone, plans to take advantage of this change in status to ignore the advice?

I am not an implementer, so I cannot answer this at all.

>>>The browser manufacturers would, I can fairly confidently state, be
>>>very keen to make this interoperable.
>>
>>The WG would be keen for the browser vendors to define what
>>"interoperable" means here. Which two parties are interoperating?
>
>I had a go at defining this more precisely, but then I saw Shawn's excellent message.

We disagree. His laundry list did not define interoperable, it was a (mostly reasonable) list of desires, plus a promise to come back later. If you want to make a run at a concise definition, that would be still be most appreciated, at least by me.

>>To start: the charter, which was widely discussed before it was
>>adopted. Next: the long history of disagreement about what kind of
>>mapping is "required" and what kind of mapping is "desirable".
>
>Is anyone arguing for the undesirability under some or any circumstances of case folding (which is the current overwhelming user expectation) and normalization?

Yes. Some argue that if the circumstance is "mandatory, fixed, and in the protocol", then it is undesirable.

>>And
>>then: what does conformance mean, as in, can a vendor make different
>>mappings if they believe that the standard got it wrong for a
>>particular set of users and/or in a particular operating
>>environment.
>
>Define "can".

"Can" means "is able".

>Saying "a vendor may choose to do something different" is no argument not to have a standard.

I wasn't arguing that. I was arguing that if you have a standard whose purpose it to give a base level of conformance, and it is in an area where there is wide agreement that local information might give them more or different actions, you need to define conformance. "MUST" means, in essence, "in order to conform to this spec, the system does this as specified".


More information about the Idna-update mailing list