On Tue, Oct 7, 2008 at 8:27 PM, Justin Karneges
<[EMAIL PROTECTED]> wrote:
> On Tuesday 07 October 2008 11:57:21 Justin Karneges wrote:
>> question whether the extra complexity of having two values instead of one
>> is worthwhile.
>
> And on that note, the best argument *I* personally see in having two values is
> with regard to usability:
>  - Most users don't want to have to specify a name for their connection.
>  - It should be possible to name a connection for those that desire to.
>  - We shouldn't use the resource for the connection name, since the resource
> is a required field and most users won't know what to put for it.
>  - Server should supply a randomly-assigned resource for all connections.
>  - Users may optionally name their connections using a separate attribute
> (one proposed way is disco identity name).
>  - Clients should not display resources of other contacts, since resources
> are useless values.  Instead, they should only display the separate name
> attribute if applicable.
>  - There ought to be a mechanism in place to prevent two named connections
> from having the same name.
>

In the whole I agree. Only I think people are being too radical :)

We are discussing 3 issues here:

1) Static resources, or dynamic?
2) User-generated, client-generated or server-generated?
3) Regardless of what we decide for the above two, people want
human-readable labels for connections

== Regarding #1: Static or dynamic? ==

I think we are (mostly) agreed that static resources can be a problem
(some users may not care, but some do, so it *is* a problem).

Note however that we have all the mechanisms in place for a safer
system already, there is no need to reinvent the wheel with a new
protocol and what-not, wait for servers to implement it, and by the
time they do, have us all change our minds again.

We need to recommend that end-user IM clients don't set a resource by
default, so that they have the server generated resource.

== Regarding #2: Who generates the resource? ==

Why do we need to set this in stone for everyone? Right now we have it
worked out perfectly. Clients who want a secure, unique, random
identifier simply ask the server to assign them one.

Clients who instead want a well-known (for routing purposes) or a
human-readable (maybe a bot would want this) resource can easily ask
for it.

No changes to the protocol required.

== Regarding #3: Human-readable identifiers for connections ==

I don't think we should be overloading disco/caps for this, it's not
what it was designed for. If people want to discuss a protocol to
solve that issue, **please start a new thread**. More threads are
good, long threads lose focus :)

Matthew.

Reply via email to