On 5/28/12 1:53 AM, Sergey Dobrov wrote:
> On 05/26/2012 01:23 AM, Peter Saint-Andre wrote:
>> On 5/23/12 1:28 AM, Sergey Dobrov wrote:
>>> On 05/23/2012 03:24 AM, Peter Saint-Andre wrote:
>>>>
>>>> But I certainly might want to receive the last published item whenever I
>>>> log in. This too seems like a setting that a dedicated microblogging
>>>> service would tune in their configuration.
>>>>
>>>
>>> Yes, it reasonable thing but do you think we should think about user
>>> subscription preferences more deeply?
>>
>> Yes, but I doubt that we could say definitively that the configuration
>> MUST or SHOULD be X for all users.
> 
> I think we can use caps for it someway. (i.e. user's resource can define
> notification behavior by setting features like +notify: +send-retracts,
> +send-last, etc)

In this area, I'd rather make some suggestions, not hard MUST or SHOULD.

>>>>> The meaning is just to provide easy way to obtain this very important
>>>>> data by just retrieving some magic constant named item. 
>>>>
>>>> We usually try to avoid magic values. :)
>>>>
>>>
>>> It's a good position and perhaps it's good enough for pubsub nodes but
>>> think about comments nodes:
>>>
>>> 1) extra node can lead to nodes names conflicts easier (again we need
>>> some magic value to construct new node name)
>>> 2) it's harder to clean up: to delete post from blog you will need three
>>> retracts which can't be done atomic.
>>
>> Yes, I see your point. I don't see a good solution right now, but I will
>> think about it some more.
> 
> Can we just add a definition of "magic" item (attached or persistent
> item) in XEP-60 and make, possibly, some special query to set/get that
> item? I think, that could solve all our problems with data sparsity. At
> the other hand, it will be easy to move from the item="0" to such magic
> item solution and we will be able to move ahead with software
> implementation until new standards will be released.

I'm still not happy with magic. However, my to-do list contains an item
for XEP-0060 revisions so perhaps we can add this feature to the open
issues list. I'll start a separate thread about it so that it doesn't
get lost in a discussion of microblogging.

> Another way is to use nodes collections but it seems to more more
> complex. For now, XEP-277 is too complicated for clients implementation,
> I think, and I'm trying to do it easier.

OK. :)

Peter

-- 
Peter Saint-Andre
https://stpeter.im/


Reply via email to