-
thilo.molitor
ralphm: like emus said
-
thilo.molitor
moparisthebest: yes dnssec is a must :)
-
moparisthebest
+1
-
jonas’
flow, making namespace prefixes (implicitly or explicitly) available to the application is exposing a detail of the serialization. making use of that detail should never be necessary, so better avoid that temptation.
-
jonas’
too many things, if exposed to namespace prefixes, make dumb decisions like matching on a prefix instead of on a URI, so better take those tools away
-
ssssssss
Vs
-
moparisthebest
jonas’: while you are correct, in fact that mistake got us here, too many things in the wild simply won't work if the stream prefix isn't 'stream'
-
singpolyma
moparisthebest: then those things are too broken to care about
-
moparisthebest
singpolyma: ejabberd is one example
-
singpolyma
I stand by my statement
-
singpolyma
If people want their stuff to work they have to make it not be hopelessly broken on the fundamentals :)
-
singpolyma
You can design around bugs sometimes, but anything that fundamental just has to work to enter consideration
-
moparisthebest
I mean someone is free to go around trying to fix everything, but untill a few years after all those patches are accepted we'll still need to use stream:stream in software we want to work in the wild, and that seems rather low on priority
-
Zash
> Interoperability Note: For historical reasons, an implementation MAY accept only the prefix 'stream' for the stream namespace (resulting in prefixed names such as <stream:stream> and <stream:features>); this specification retains that allowance from [RFC3920] for the purpose of backward compatibility.
-
Zash
Nice roundabout way of saying MUST send `<stream:stream xmlns:stream="...">`
-
moparisthebest
Yep, don't fight the man
-
Zash
That and the separate namespaces for c2s and s2s (and components and BOSH and WS...) would have been nice to fix for XMPP 2.0 at some point in the distant future.
-
jonas’
moparisthebest: as I said yesterday: an XML generator should allow *defining* prefixes precisely for such cases. a parser should never expose them to prevent such things from happening in the first place
-
Zash
So all the programmer should be dealing with is (namespace, name) tuples
-
singpolyma
Zash: or ideally just name, which includes namespace
-
Zash
I would consider that an implementation detail.
-
jonas’
exactly
-
moparisthebest
jonas’: yes, nice