Re: [xmpp] [precis] WGLC Comments on 6122bis

["Ben Campbell" <[e-mail]>]

On 10 Mar 2015, at 18:13, Peter Saint-Andre - &yet wrote:

> On 3/10/15 4:38 PM, Ben Campbell wrote:
>> BTW, I reviewed version 19, quite by accident :-)
>>
>> On 10 Mar 2015, at 17:30, Ben Campbell wrote:
>>
>>> (as individual)
>>>
>>> Hi,
>>>
>>> This is a very well written draft. It's easy to understand given the
>>> complexity of the material. I support it's publication.
>>>
>>> I have a few minor comments:
>>>
>>> -- section 3.2, paragraph 6:
>>>
>>> Is there a reason to avoid a reference for IDNA2008?
>
> Well, IDNA2008 is collectively RFCs 5890-5894 and we do point to 
> several of those specs in §3.2. :-)

I guess I should have said "... here", but needing 5 references probably 
argues against doing it again. :-) It just sort of looked strange that 
it was missing while the next phrase had a reference.

>
>>> -- 3.3, implementation note:
>>>
>>> Are there any practical consequences for the implementor? Are there
>>> potential conflicts where the XMPP implementation correctly forms a
>>> Localpart, but it contains an identifier that is interpreted
>>> incorrectly by some SASL mechanism?
>
> Re-using the UsernameCaseMapped profile from 
> draft-ietf-precis-saslprepbis should help in this regard. However, as 
> noted, some SASL mechanisms might not be upgraded quickly and thus 
> would still use SASLprep (RFC 4013). The differences are explained a 
> bit more in Appendix A of draft-ietf-precis-saslprepbis - as I see it, 
> the only semi-major issue is that certain characters that were "mapped 
> to nothing" in RFC 4013 are simply disallowed by the 
> UsernameCaseMapped profile that we re-use in 6122bis.

So an identifier created with UsernameCaseMapped should be fine, but an 
identifier created somewhere else in sasl might not be legal? That 
_seems_ unlikely to be a problem...

It might not hurt to mention the saslprepbis appendix in the note.

>
>>> -- 3.3.1, implementation note:
>>>
>>> I have mixed feelings about XEP-0106 being an informational 
>>> reference
>>> (using the standard of "need to read to understand/implement this
>>> document"). Even if an implementation chooses not to create JIDs 
>>> with
>>> escaped characters, it had to be prepared to receive them from
>>> somewhere else, doesn't it?
>
> Well, the alternative is to reject the JID - but since the escaping 
> mechanism in XEP-0106 is really a matter of presentation, it's 
> something that affects clients more than servers anyway (and the JID 
> escaping would be applied before PRECIS-related processing happens 
> anyway).

So a server would just operate on the escaped string without worrying 
about the fact. Got it. OTOH, this draft applies to clients, too. I 
suppose a client might display the raw string without unescaping, but 
that would be ugly.

OTOH, that goes beyond the point of _this_ draft, so I'm okay leaving it 
as informational.

[...]

>>> -- 8:
>>>
>>> While I don't object to the approach of the section, I think there's
>>> some risk of confusion about which text is authoritative from a 2119
>>> perspective. I think it might be worth noting that the authoritative
>>> text is in the referenced sections, and only summarized here for
>>> convenience.  (It only really matters if they conflict, but 
>>> redundant
>>> normative text makes life harder for future updates.)
>
> I see your point: it's never a great idea to say the same thing in two 
> places.
>
> We do say:
>
> This section describes a protocol feature set that summarizes the
> conformance requirements of this specification....
>
> We could add a sentence like:
>
> The summary is purely informational and does not override any of the
> more detailed descriptions in the body of this specification.


How about adding something more specific. For example, something to the 
effect of the following (to either the current text or with your 
proposed addition):

"[RFC2119] keywords in this section are intended to describe the 
referenced normative text."

(A careful reader will figure out the right thing without it. But I'm 
not worried about _careful_ readers ;-) )