[Standards] publish+configure again

Ian Paterson ian.paterson at clientside.co.uk
Fri Mar 30 13:49:49 UTC 2007


Jacek Konieczny wrote:
> On Fri, Mar 30, 2007 at 01:36:39PM +0100, Ian Paterson wrote:
>   
>> p+c will make it much less likely, but *not* impossible, that clients 
>> will inadvertently interfere with one another.
>>     
>
> IMHO using p+c may make this problem even more more likely. When a node
> is configured only when created and later only if user asks, things are
> predictable -- if it worked, it will still work. If any client would
> change node settings on each item publish, then any client would change
> settings of the node, whenever it makes sense or not.

Disallowing p+c isn't going to stop clients doing that. XEPs need to 
make the values of some config parameters mandatory. If clients are 
going to use p+c for parameters whose values are permitted to vary, then 
they must ensure the user understands what the client is doing 
(something for the PEP Implementation Notes).

> Let the node
> configuration be an account setting, not a client setting (most probably
> hard-coded in most clients).
>   

We've already been around and around this argument. You can always 
invent scenarios where it is better not to use p+c... but we can always 
invent scenarios where it is better to use it. That is why p+c is optional.

FYI, in some cases, like geoloc, IMHO p+c should *not* be used. In other 
instances, like "Home Address", p+c makes things easier (and offers 
atomicity - which may eventually prove valuable in some cases).

- Ian




More information about the Standards mailing list