Re: [salud] Christer's review of draft-ietf-salud-alert-info-urns-09

[Christer Holmberg <[e-mail]>]

Hi,

>> BTW, the following sentence is unclear to me:
>> 
>>    "In practice, this specification extends Alert-Info in that it will
>>    cause the use of a new class of URIs and the use of multiple URIs."
>> 
>> What does it mean? 
>
> The idea is that, while 3261 syntactically and semantically *allows* non-HTTP URIs in Alert-Info, nobody ever used non-HTTP URIs in Alert-Info, and nobody implemented an interpretation of those URIs.
> Thus, even though we are not changing what is officially permitted, implementers will have to change their code because the new *practice* will be an extension of the old practice.  (And both the new practice and the old practice are a small subset of what is permitted by 3261.)

Well, then you should use different wording. Because, the draft does not extend Alert-Info, it simply describes the usage of a new URN in the header field.


>>>> Q9:
>>>> -----
>>>> 
>>>> Section 4 defines the ABNF for the URN, but [there] is no text on 
>>>> how to "map" it into the Alert-Info header field ABNF:
>>>> 
>>>> For example, I assume that the URN is encoded using the opaque-part 
>>>> format of the absoluteURI, and that the scheme value is "urn". I 
>>>> think it would be good to indicate that.
>>>
>>> Actually, we're depending on the fact (unstated in RFC 3261) that 
>>> any "absolute" URI (per RFC 3986) is allowed as a header field value 
>>> in Alert-Info, and that the URNs we are defining conform to the 
>>> absolute URI syntax.  If the syntax for <absoluteURI> in RFC 3261 
>>> didn't allow all absolute URIs, we'd have to amend RFC 3261.
>> 
>> There are two different "structures" for absolute URI, and the URNs 
>> only fit into the opaque-part structure.
>> 
>> In addition, as the URI requires a scheme value, I think we should 
>> explicitly say what it is.
>
> I guess my point is that the 3261 ABNF already allows all alert URNs to appear in Alert-Info, and there is no need to specify exactly how alert URNs are compatible with the 3261 ABNF.  Anyone who cares exactly how this draft is directly compatible with the ABNF can see that by looking at the ABNF.
>
> Have there been other situations where this sort of syntax explication has been provided?

I don't know - I am just commenting on how I think the draft could from my perspective be improved :)

Regards,

Christer