Hi, 

> >> we should be ready for a transition to a new session description 
> >> protocol, which includes experimental networks using their 
> own stuff, 
> >> and people seem to agree that alternative is the way to 
> go. That is 
> >> why it would be great if we could mandate its support already now.
> >>
> >> The more we wait, the more trouble we could have in the 
> future to use 
> >> it.
> > 
> > Yes.
> > 
> > But, at the same time we need to try to be backward compatible, and 
> > take into consideration existing implementations which may support 
> > mixed but not alternative. If those implementations reject 
> alternative 
> > I think our spec should not make them "not compliant".
> 
>The MIME spec already says that if you don't understand 
>multipart/X then you should treat it as if it were 
>multipart/mixed. So at a minimum we should expect 
>implementations to do that much with alternative.

Well, assuming we don't think THAT could cause problems in the future...

But, since that is generic specific MIME behavior I think we'll have to
live with it.

But, I STILL think it's good to give an implementation to option to
reject alternative - at least if we find out that is what existing
implementations are doing.
  
>If we are going to make more accommodation than that based on 
>what is in the wild then I suggest we should take a survey 
>first to find out what we really need to cope with. We should 
>be able to get some good data from SipIT. Maybe we should be 
>trying to come up with a list of specific questions, or maybe 
>we can have an essay question on the survey to get a 
>description of exactly what is/isn't supported.

My understanding is based on my SIPit experience, and what I have
implemented myself.

But, I have to admit I haven't been to SIPit for a while, so someone
else (Robert?) may have better information.

>I think we will have a pretty good chance of seeing support 
>in the future if we can tell implementors what they must do. 

It's not always the implementors we have to tell - often it's the people
paying the implementors :)

>I did specifications for multipart support in a stack a few 
>years ago (which is when I discovered all these open issues), 
>and I had to make up what was done based on speculation about 
>what would be used. Even hard things can be straightforward 
>to implement if the requirements are clear.

I am not saying that it is difficult to implement alternative, but I
doubt many have done it.

But, it would be interesting to know what implementations do: 

1. do they treat alternative as mixed? 
2. do they reject the request?
3. do they support alternative? 
4. do they discard the message body?

I think our spec should allow options 1, 2 and 3.

Regards,

Christer











> >>>  
> >>> Hi,
> >>>
> >>>>>> OPEN ISSUE 1: do we need to say something about other types?  
> >>>>>> Related (rfc2387), parallel, digest, external-body.
> >>>>> I don't know much about these. I quickly scanned 2387 and
> >> see that
> >>>>> it has some special rules re processing
> >> Content-Disposition of the
> >>>>> contained (related) parts. I *think* those are consistent
> >> with what
> >>>>> we are doing.
> >>>> I have added the following text to Section 5:
> >>>>
> >>>>    "This specification recommends support for 
> 'multipart/mixed' and
> >>>>    'multipart/alternative'.  At present, there are no SIP
> >> extensions
> >>>>    that use different 'multipart' subtypes such as parallel 
> >>>>    [RFC2046], digest [RFC2046], or related [RFC2387].  If such 
> >>>>    extensions were to be defined in the future, their 
> authors would
> >>> need to make sure
> >>>>    (e.g., by using an option-tag or by other means) that entities
> >>>>    receiving those 'multipart' subtypes were able to
> >> process them.  As
> >>>>    stated earlier, UAs treat unknown 'multipart' subtypes as 
> >>>>    'multipart/mixed'."
> >>> I think the text is good.
> >>>
> >>> But, I wonder whether we should alternative to the list of
> >> "different
> >>> multipart subtypes". At present there are no SIP extensions using 
> >>> alternative either, so...
> >>>  
> >>> Or, at least we could leave support of alternative open for the 
> >>> moment, until we get more information about how it's 
> supposed to be 
> >>> used in the first place (regarding identical content type
> >> values etc).
> >>>
> >>>>>> OPEN ISSUE 2: we know that we do not want two SDPs in a
> >> 'multipart/
> >>>>>> alternative', but is this valid generally with any 
> content type?
> >>>>>> Would it be possible to provide two alternative body 
> parts using 
> >>>>>> the same format and, thus, the same content type but in, say, 
> >>>>>> different languages?
> >>>>> Its my understanding that the distinction is based on
> >> which can be
> >>>>> understood, relative to Content-Type. It isn't apparent
> >> to me that
> >>>>> making this decision based on other attributes is valid.
> >>>>> For one thing the parts are supposed to be ordered by 
> increasing 
> >>>>> richness. If they differed by language this wouldn't be true.
> >>>>>
> >>>>> So I think what you have is ok.
> >>>> I double check with an email expert to make sure we get 
> this right.
> >>>
> >>>>> Another thing in Section 3: You have used SHOULD in this
> >> section. I
> >>>>> thought we wanted to change this to MUST. Did we?
> >>>> You refer to the following statement, right?
> >>>>
> >>>>    "In particular, UAs SHOULD support the 'multipart/mixed'
> >>>>     and 'multipart/alternative' MIME types."
> >>>>
> >>>> If people agree that this should be a MUST, I will be
> >> happy to change
> >>>> it.
> >>> Again, I think we should wait a little regarding alternative.
> >>>
> >>> Regards,
> >>>
> >>> Christer
> >>
> > 
> 


_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to