[jdev] Short thoughts about DNS/SRV, Unicode, Stanza namespace prefix

[Rene Treffer]

:: 1 - DNS/SRV

Priorites and weight should help to balance hosts. However I'm not yet
sure how to combine this with unreliable networks that randomly break
(think mobile). Ideas / recommendations would be welcome.

:: 2 - Encoding

Character Encoding: It is not clear if a client/server may send a UTF-8
BOM (byte order marker). This should be clarified. I'm not in favor of
either way, but it should be pointed out.
Unicode: It is not clear how a stream should handle invalid (ill) utf-8
sequences.

:: 3 - XML Stanza namespaces

This is about the section Peter outlined: The stream stanzas must not be
prefixed. I learned this by testing against gtalk (which happily
violates this) and trying to log into an ejabberd afterwards....

I'm pretty much in favor of relaxing this rule. A prefixed stanza can be
fully analyzed on its own without trouble. It's a "detachable" version.
(I'm actually using this to feed a smack parser. it's way more stable
than crashing the connection, working on a shared/direct xml parser is
risky.). I don't really see a reason why this rule should be stronger
than "should not" (at least in the long run).

Regards,
  Rene Treffer