[standards-jig] NEW: Message Delivery Semantics (JEP-0079)

Matthew A. Miller linuxwolf at outer-planes.no-ip.com
Thu Apr 17 16:35:08 UTC 2003


On Wed, 2003-04-16 at 17:55, David Waite wrote:
> A few brief comments:
> 
> - The resource is defined in the XMPP spec as being opaque - I should 
> not depend on a prefix/suffix actually meaning anything, or structured 
> in any style. (For an example in terms of another protocol: I shouldn't 
> assume http://example.com/cgi-bin/ refers to the parent document of 
> http://example.com/cgi-bin/form.pl)

The intention of this JEP was not to add meaning to the resource, but to
perform substring matches against the resource.  Would adding additional
rules, specifically "suffix-based" and "anyfix-based", help prevent this
violation?

So far, there does not seem to be objections by anyone else, but then
only three people have voiced their opinions.

> - Section 3 (Implementation Notes) is very odd; a client cannot perform 
> any of the routing semantics, and the expiry semantics are not defined 
> in application terms/client UI terms. Does it exire if not delivered by 
> X time or if not displayed/processed by X time?

The intention here was that clients MAY wish to extend the behavior into
their user experience, by dropping/failing the message if it had not
been displayed within the expiration timeframe.  For servers (that
understand this JEP), this values always denotes delivery time.

I can make that section clearer on this point, or remove it altogether.

> - Do you see additional rules being defined later? Perhaps the rules 
> supported should be reported in a disco query response? A client would 
> need to do a disco query of the remote server before requesting any of 
> these delivery semantics anyways.

At this point in time, I don't see more rules, but others do.  Making
the specific rules supported somehow discoverable would be nice, but I'm
not sure how exactly to incorporate that into disco.

I suppose one method would be to have a <feature/> for each condition
and action, in the following general forms?
conditions ==
"http://jabber.org/protocol/msg-delivery?condition=<condition name>"
actions == "http://jabber.org/protocol/msg-delivery?action=<action
name>"

The presence of the feature "http://jabber.org/protocol/msg-delivery"
would imply support for only those conditions and actions defined in
this JEP.

> - Would a client ever want to request certain rules be used when 
> responding to a message, or within a thread?

The semantics are only intended for the sender to make.  Is there a
use-case that the recipient would want senders to use certain semantics?

> - Scoping larger than a single message (such as a whole thread) opens a 
> whole can of worms. Does a by-thread scope apply to all resources or 
> just the current one? How does the server prevent a user from doing a 
> DoS by adding a few million thread IDs and eating up all memory? When 
> does the thread scope end (when the local user goes offline? when the 
> remote user goes offline?)  I'm not saying whether the bandwidth it 
> would save is worth the the complexity in the proposal and 
> implementations or not, just that it needs to be taken into account :-)

The more I think of thread-level scope, the more I can see problems. 
These semantics need to be distributed to all routes on the chain
between any two entities, and can have a larger scope of problems. 
While today that would mean at most two servers, the reduced stanza size
this provides as a convenience, it creates a different, widespread
problem.

Thanks for convincing me that thread-level scope is bad (-:

> -David Waite
> 
> Peter Saint-Andre wrote:
> 
> >Matthew Miller has submitted a JEP for message delivery semantics. This
> >JEP addresses the recent discussions on this list. You can find it here:
> >
> >http://www.jabber.org/jeps/jep-0079.html
> >
> >Enjoy!
> >
> >Peter
> >
> >  
> >
> 
> _______________________________________________
> Standards-JIG mailing list
> Standards-JIG at jabber.org
> http://mailman.jabber.org/listinfo/standards-jig
-- 

Matt "linuxwolf" Miller
JID:	linuxwolf at outer-planes.net
E-MAIL:	linuxwolf at outer-planes.net

- Got "JABBER"? (http://www.jabber.org/)




More information about the Standards mailing list