[Standards-JIG] UPDATED: JEP-0030 (Service Discovery)

Richard Dobson richard at dobson-i.net
Tue Mar 16 17:04:00 UTC 2004


> Of course we don't want two discovery methods. I was just offering a
humble
> solution to a weakness of JEP-0030 that I perceive (and Jacek touched on).
> i.e. that Disco scales very badly when the client requires just a little
> piece of information about hundreds or thousands of items. I would not
> advocate that JEP writers allow the 'verbose' option in other
circumstances.

Yup I can certainly see its a problem, but is was seen as a reasonable trade
off and was known about from the start, its not something that has only
cropped up recently. The problem with your solution is that it goes against
the primary reason disco was developed to replace browse in the first place,
i.e. the parent entity shouldnt be supplying information about its child
items, you must request further information from the child to find out what
it is etc, the main reason IMO that this was done is otherwise you need to
then somehow have communication between the child item and its parent so the
child can ensure the information the parent has about it is correct and up
to date which becomes a complete nightmare keeping things in sync. For your
solution to work parent items would again have to know information about
their child items which takes us right back to the situation we had with
jabber:iq:browse, disco was designed to be simple (your suggestion makes it
much more complex as it has to be able to work in two completely different
ways) its probably best not to radically alter the way it works as you
suggest even if its just an extension.

> I *honestly* bow to your superior experience (and that of many other
people
> on this list). But I obviously still don't understand the serious harm in
my
> suggestion... So thank you in advance for your patience. ;)

I hope the above helps clarify things :)

Richard




More information about the Standards mailing list