Re: [xmpp] Barry Leiba's No Objection on draft-ietf-xmpp-posh-04: (with COMMENT)

[Peter Saint-Andre <[e-mail]>]

Hi Mark, thanks for sharing your perspective. I'll give it more thought tomorrow. 

Peter

Sent from mobile, might be terse 

> On Aug 3, 2015, at 6:10 PM, Mark Nottingham <[e-mail]> wrote:
> 
> Hi Peter,
> 
> As I discussed with Barry, I don't have any strong guidance here; I agree that his approach is probably the way I'd go with if I designed it myself, but I agree it's highly dependent on the situation. With my DE hat on, I'm neutral on this. 
> 
> Cheers,
> 
> 
>> On 4 Aug 2015, at 9:31 am, Peter Saint-Andre - &yet <[e-mail]> wrote:
>> 
>> Adding Mark to this thread, with pointers to get him up to speed...
>> 
>> On 7/31/15 7:50 PM, Barry Leiba wrote:
>>>>> If you like, it could even do POSH service names and POSH format
>>>>> names, and specify ".well-known/posh/<servicename>/<formatname>",
>>>> 
>>>> Currently it's the <servicename> field that we're most interested in.
>>> 
>>> Right... and that's what I'm suggesting an FCFS registry for.  You
>>> register "posh" in .well-known, and you create your own FCFS registry
>>> for service names, and if you don't care about the format as a
>>> separate thing, you just register "spice.json" (and so on) in your
>>> FCFS registry.  That way, Mark doesn't get involved in approving
>>> "posh.x" and "posh.y" and "posh.z", when Mark has no idea of what to
>>> say about posh service names (or seedy ones, for that matter).
>>> 
>>>> Perhaps it makes sense to talk about it again with Mark?
>>> 
>>> Sounds like a plan.
>> 
>> Hi Mark,
>> 
>> During IESG review of draft-ietf-xmpp-posh, Barry raised a question about the .well-known registration, which you and Matt Miller and I talked about ages ago. Here are some relevant readings:
>> 
>> https://tools.ietf.org/html/draft-ietf-xmpp-posh-04#section-8
>> 
>> https://mailarchive.ietf.org/arch/msg/xmpp/Vcikwhh1Fln6Z7WfKlb0mCarFDw
>> 
>> https://mailarchive.ietf.org/arch/msg/xmpp/5bRV3Ob2ugDkL2E7XFrCL-eKLFE
>> 
>> https://mailarchive.ietf.org/arch/msg/xmpp/xw_-TOVxkQJWqquZe4UH2yDJKAg
>> 
>> Barry's proposed approach is more elegant. However, my feeling is that POSH is essentially a one-off workaround for use in XMPP until DNSSEC and DANE are more widely deployed. Although we've tried to design it in such a way that it *could* be used by other application protocols, I doubt that folks in those communities will use POSH unless Matt and I start actively promoting it (and even that is only a necessary condition, not a sufficent one). Because I don't think that we'll see additional .well-known registrations related to POSH, setting up a separate POSH registry to supplement a "posh" entry in the well-known URIs registry feels like overkill to me given the burden on IANA (I'm not a fan of one-entry registries).
>> 
>> That said, people have been wrong before about the popularity of technologies defined in RFCs and it's possible that POSH could become more popular, in which case I'd lean more strongly toward the approach that Barry has outlined.
>> 
>> As the designated expert for the well-known URIs registry, do you have any suggestions or preferences on how to proceed?
>> 
>> Thanks!
>> 
>> Peter
> 
> --
> Mark Nottingham   https://www.mnot.net/
>