On 6/23/13 8:27 PM, Peter Saint-Andre (psaintan) wrote:
> Hi Brian, thanks for the review.

Here is a provisional diff that addresses your feedback:

https://github.com/emcho/cusax/commit/1c5e9d681748685360cc6677ea2e369b352e66db#draft-ivov-xmpp-cusax-07.xml

> 
> On Jun 23, 2013, at 8:01 PM, Brian E Carpenter <[email protected]>
>  wrote:
> 
>> I am the assigned Gen-ART reviewer for this draft. For background on
>> Gen-ART, please see the FAQ at
>> <http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
>>
>> Please resolve these comments along with any other Last Call comments
>> you may receive.
>>
>> Document: draft-ivov-xmpp-cusax-06.txt
>> Reviewer: Brian Carpenter
>> Review Date: 2013-06-24
>> IETF LC End Date: 2013-07-16
>> IESG Telechat date:
>>
>> Summary:  Almost ready
>> --------
>>
>> Comment:
>> --------
>>
>> I have some issues, but I don't know whether to classify any of them as
>> major.
>>
>> Issues:
>> -------
>>
>> In Introduction:
>>
>> "  As a result, a number of adopters have found themselves needing
>>   features that are not offered by any single-protocol solution, but
>>   that separately exist in SIP and XMPP implementations.  The idea of
>>   seamlessly using both protocols together would hence often appeal to
>>   service providers. "
>>
>> A few paragraphs later, you discuss the case of an end user who would like
>> to use SIP and XMPP together with *different* service providers.
> 
> Although that's out of scope for this document. We might address that 
> scenario in a future document.
> 
>>  I would suggest
>> ending this sentence with "appeal to service providers and users."
> 
> Good point.
>>
>> "  Finally, this document makes a further simplifying assumption by
>>   discussing only the use of a single client, not use of and
>>   coordination among multiple endpoints controlled by the same user
>>   (e.g., user agents running simultaneously on a laptop computer,
>>   tablet, and mobile phone)."
>>
>> Hmm. Isn't that the normal case today, and your simplifying assumption
>> the exception? It's certainly extremely annoying to have different
>> contacts lists on different devices, for example. This seems like a
>> big gap in the model, with no hints on how it might be filled. (As big
>> as the gap between POP3 and IMAP, in some ways.)
> 
> At least in XMPP, you'd have the same contact list on all of those devices. 
> But we can word this more carefully.
> 
>>
>> In Client Bootstrap:
>>
>> "  While it should be possible for CUSAX users to manually configure
>>   their separate SIP and XMPP accounts, service providers offering
>>   CUSAX services to users of dual-stack SIP/XMPP clients ought to
>>   provide means of online provisioning,..."
>>
>> 1. I would anyway expect my CUSAX client to come with a configuration
>> wizard, including a path to the online provisioning if available.
> 
> True.
>>
>> 2. Is there any reason why SIP service providers and XMPP service
>> providers shouldn't individually provide on-line provisioning?
> 
> No.
> 
> (Something like the technology proposed at the aggsrv BoF in Orlando might 
> help, too, but that's just a gleam in the eye at this point.)
> 
>>  You're
>> describing something close to a captive-customer scenario,
> 
> We are indeed assuming that a dual-stack user agent would interact with a 
> combined service (i.e., CUSAX on the client side and CUSAX on the server 
> side).  I am not sure that a customer of such a service is a captive.
> 
>> rather than
>> encouraging an open approach to provisioning.
> 
> As far as I can see, no open provisioning protocol is available yet. The 
> aforementioned aggsrv approach might fit the bill. It's too early to say, but 
> we can at least mention the possibility that such a technology might be 
> developed in the future.
> 
>> The CUSAX client would
>> in any case have to deal with any inconsistencies in provisioning.
> 
> I'm not sure what actionable advice we can provide here, but I'll discuss 
> with with my co-authors.
>>
>> In Server-Side Setup:
>>
>> "  In order for CUSAX to function properly, XMPP service administrators
>>   should make sure that at least one of the vCard [RFC6350] "tel"
>>   fields for each contact is properly populated with a SIP URI or a
>>   phone number when an XMPP protocol for vCard storage is used (e.g.,..."
>>
>> How can they do that, given that users are normally responsible for
>> maintaining their contacts lists?
> 
> When a user uploads their vCard (or, say, modifies it in a web interface), 
> the server could apply automated checks or make the "tel" field mandatory.
>>
>> In Summary of Suggested Practices:
>>
>> "  1.   By default, prefer SIP for audio and video, and XMPP for
>>        messaging and presence."
>>
>> At the beginning of the Operation section, this seems to be stated as
>> a rule, not as a default (" Audio/video features however, are disabled
>> in the XMPP stack..."). Which is it?
> 
> I'm not seeing a huge difference here -- the XMPP stack in the CUSAX user 
> agent might have audio/video capabilities, but typically those would be 
> disabled (disabled by default, but the user might be allowed to turn them on 
> for certain kinds of contacts, etc.).
> 
> I suggest this change in the Operation section.
> 
> OLD
>    Audio/video features however, are
>    disabled in the XMPP stack, so any form of communication based on
>    these features (e.g. direct calls, conferences, desktop streaming,
>    etc.) will happen over SIP.
> 
> NEW
>    Audio/video features however, would typically be
>    disabled in the XMPP stack, so any form of communication based on
>    these features (e.g. direct calls, conferences, desktop streaming,
>    etc.) would happen over SIP.
> 
>>
>> In Security Considerations:
>>
>> "  ... a CUSAX client might
>>   successfully negotiate Transport Layer Security (TLS) [RFC5246] when
>>   connecting to the XMPP aspect of the service but not when connecting
>>   to the SIP aspect.  Such mismatches could introduce the possibility
>>   of downgrade attacks."
>>
>> I'd say *would* introduce the possibility.
> 
> Yes, that's more accurate.
> 
>>  It would seem possible for a
>> bad actor to pick up authentication data from the insecure service and
>> exploit it to attack the secured service. Therefore,
>>
>> "  User agent developers and service providers
>>   ought to ensure that such mismatches are avoided as much as possible."
>>
>> seems a bit weak. Shouldn't the client also be *required* to alert the user
>> that the session as a whole is not secured?
>>
> 
> IMHO that would be advisable. Good catch.
> 
> Peter
> 
_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art

Reply via email to