jdev - 2021-10-17

does a element name like <stream:stream> in xml has a special meaning because of the ":" or is this just a element name like any other and ":" has no special meaning

i tried to find it in the xml spec, but this thing is unreadable for me

lovetox, IIRC it has no special meaning if there are no XML namespaces

I think https://www.w3.org/TR/xml-names/ is what gives ":" in xml element names a special semantic

So as soon as you use namespaces, which is probably true for most practical usages of XML, it has a special meaning

Including in XMPP

lovetox, but I am curious, as it look like there is some more backstory to this question. if so, would you mind elaborating a bit? :)

the backstory is i try to rewrite the nbxmpp streamparser and now i have to understand this :)

lovetox, I'd hoped that python has a good xml pull parser library that you could re-use?

otherwise I feel like you will hit a few xml trapdoors and probably re-work your implementation 2-3 times :)

but you will learn a lot about XML in the process, so not sure if this will be a loss, just be prepared

500% use an existing parser

Please please please

make sure to look at existing APIs. I think Java's StAX is a very good example of an XML stream parser

the standard library parser is good enough for that

it is based on Expat and does SAX

please use that.

expat is a great choice

just remember to disable all the security holes

it can be used non-blockingly. If you need proof, look at aioxmpp.

at least that's easier than reinventing the security holes yourself

calm down, i use expat

jonas’, isn't SAX a push parser? when can those not be used non-blockingly?

still need to understand xml namespaces

I think recent expat versions have more sensible defaults, though forbidden XMPP stuff will still need handling

lovetox, do you? the parser should do all that for you.

(and expat does if you ask nicely)

(and by extension, the python standard library xml module does)

I think you need to understand XML when parsing XML, and even if not, it can't hurt ;)

https://github.com/horazont/aioxmpp/blob/devel/aioxmpp/xml.py#L818-L1111 explicitly waiving the infectiousness of the *GPL for you, lovetox, if you need/use that as inspiration.

jonas’, if i search for nx prefix in that code, i find a lot of matches

thanks flow, that document did help to understand prefixes :)

nx?

*ns

regarding your comment, that its not necessary to understand and python xml libs do it for you

The summary is that the element called "foo" in one namespace is not the same as the element called "foo" in another namespace. A namespace is identified by a string, usually found in the 'xmlns' attribute. Prefixes are simply a more concise replacement for adding 'xmlns' attributes everywhere (you declare them once and use them as many times as you want)

This looks like a sane explanation: http://tech.usgin.org/content/xml-namespaces-and-prefixes

XMPP is heavily based on XML namespaces, so understanding them is not really optional for an XMPP dev. But a good parser (such as expat) will save you from having to handle all the rules (such as inheritance) yourself, and it will tell you what namespace every element is in (regardless of whether it used a prefix or an xmlns attribute)

yes that would be nice

i just tested expat

and it does not handle it

at least the expat api that is accessible via python

hm altough some examples show it does, maybe i need to investigate this more

Expat can optionally do XML namespace processing for you, enabled by providing a value for namespace_separator

oh i did specify that :)

lovetox, check that: https://github.com/horazont/aioxmpp/blob/devel/aioxmpp/xml.py#L1106-L1111

jonas’, i use currently xml.parsers.expat.ParserCreate

is there any benefit of using the sax module?

less boiler plate to write I guess

I strongly suggest to go with the high-level SAX API, it saves you worrying about namespace prefixes and stuff.

ok i look into it

n

and all the prefix stuff is just to save on size? it does not add some new feature to xml? so if a client does not use ns prefixes at all its perfectly fine?

prefix stuff is generally not used in XMPP, apart from for <stream:stream> and Dialback

Does it not add the ability to namespace attributes? I.e how else would you say `<e xmlns:foo="urn:example:ns" foo:bar="blah"/>` ?

(Which is also pretty rare in XMPP)

yeah it does, forgot about the attribute stuff

its weird in one example the stream prefix is used

in others not in the rfc

Hm?

Like `<stream ....` ? Probably just to make the examples more compact

for example here its used

https://datatracker.ietf.org/doc/html/rfc6120#section-9.1.2

https://datatracker.ietf.org/doc/html/rfc6120#section-9.2.1

here than suddenly missing from the features

🤷️

i think they want to make a point that it can be used

But doesn't it also say that you should encode the stream as `<stream:stream ...>` ?

https://datatracker.ietf.org/doc/html/rfc6120#section-4.8.5

oh .. wait

its not optional

if we want to define a default namespace of jabber:client in <stream>

you need a prefix if <stream> should not have jabber:client

is this example not wrong here

S2: <stream from='example.net' id='hTiXkW+ih9k2SqdGkk/AZi0OJ/Q=' to='im.example.com' version='1.0' xmlns='http://etherx.jabber.org/streams'>

it does not define a default namespace of jabber:client or server

`<stream xmlns="http://etherx.jabber.org/streams"><message xmlns="jabber:client"/></stream>` == `<stream:stream xmlns:stream="http://etherx.jabber.org/streams" xmlns="jabber:client"><message/></stream:stream>` ✏

hm yeah

but without defining it on the stream element, this would indeed be very wasteful

So if you could be certain that the other end of the stream understands namespaces and prefixes fully, you could cut down on repeated namespace declaration for common things. Like imagine `<sm:r/>` instead of `<r xmlns="urn:xmpp:sm:3"/>`

test

hmm

This was a triumph! I'm making a note here: HUGE SUCCESS!