[Jabber-IETF] Agenda items
iain.shigeoka at messaginglogic.com
Wed Oct 9 23:20:28 CDT 2002
On 10/9/02 5:28 PM, "Robert Norris" <rob at cataclysm.cx> wrote:
>> I see what you're saying. I'm a bit concerned about the flexibility of XML
>> in a wire protocol. As in this example, there are multiple XML compliant
>> ways of doing the same thing. The KISS principle makes me a bit
>> uncomfortable with this. I think I'd really prefer to see one particular
>> XMPP valid format for the packets and namespaces unless someone can show a
>> concrete benefit for allowing this flexibility.
> It really depends on how the stream implementation is written. I know
> from experience that it is fairly straightforward to support the full
> flexibility of XML in the stream.
> I would argue the other way - we're not likely to be able to think of
> all the possible ways in which someone might want to use a stream, so
> lets not set arbitrary limits which we might have difficulty changing
Yes but it still violates the KISS principal and recent engineering theories
like XP: don't build it until you need it. We don't need it yet. I would
prefer not to design for theoretical future needs. Simple simple simple...
:) I'd rather add it later when a real need arises rather than try to guess
what we'll need later and complicate things in the meantime.
>> I know I keep harping on the same issue like a broken record... *sigh* I
>> just have a bad feeling about the number of possible valid variants coming
>> back to bite us. Part of the simplicity of the current Jabber
>> implementations is that jabberd is essentially the rosetta stone for
>> clients. Practically speaking, there is only one form of tag names,
>> prefixes, and namespaces in the wild. I'm really worried about a player that
>> likes to "embrace and extend" coming along (if we allow full XML
>> extensibility in the core) and exploiting XMPP flexibility to lock everyone
>> out while claiming compatibility and spreading FUD...
> I fail to see how that can happen. If we have a number of discrete data
> types, each in their own namespace, then surely the worst they can do is
> add one more?
Some companies have been quite successful at taking something standard and
supposedly simple and making it really hard to deal with unless you have
their stack. I would take as example the theoretically simple HTML page, and
what is produced by some/most word processing applications. It is
conceivable for someone to build implementations that can properly display
the simple HTML page, but how many would attempt to hack that today? I don't
want Jabber software to become something only large companies can build.
More information about the xmppwg