Fix SIP URI support for `urllib.parse.url{parse,split}` (RFC 3261)

[codingjoe]

Bug report

Bug description:

urllib.parse currently supports the SIP URI scheme. However, it mislabels the netloc as the path.

>>> import urllib.parse
>>> urllib.parse.urlparse('sip:user:password@host:port;uri-parameters?headers')
ParseResult(scheme='sip', netloc='', path='user:password@host:port', params='uri-parameters', query='headers', fragment='')

RFC 3261 defines the sip URI without slashes, like so:

sip:user:password@host:port;uri-parameters?headers

The correct ParseResult would be:

>>> urllib.parse.urlparse('sip:user:password@host:port;uri-parameters?headers')
ParseResult(scheme='sip', netloc='user:password@host:port', path='', params='uri-parameters', query='headers', fragment='')

CPython versions tested on:

3.14

Operating systems tested on:

macOS

[picnixz]

FTR, this also affects urllib.parse.urlsplit so the problem isn't the "obsolete" RFCs urlparse is based on.

[picnixz]

I'll consider this a feature. I am not entirely sure we want to support schemes without slashes by default. Do other (more common) protocols define an alternative syntax without slashes?

[codingjoe]

I'll consider this a feature. I am not entirely sure we want to support schemes without slashes by default. Do other (more common) protocols define an alternative syntax without slashes?

I saw multiple references to SIP in the code base, which is why I reported this as a bug.

On that note: SIP is no obscure protocol. It powers most of the call routing on the planet and is equal to HTTP in most regards including its obscure past. Which yes, uses URIs without a double slash. 😅

[picnixz]

I'm not saying SIP shouldn't be supported. SIP is documented as being supported. What I am unsure is slash-less SIP. In the docs, we assume that schemes are followed by :// I think.

[picnixz]

Or maybe there is no slashful SIP? if this is the case, it's indeed a bug!

[codingjoe]

Or maybe there is no slashful SIP? if this is the case, it's indeed a bug!

There isn't a slashed version. Same like as the tel:-scheme.

[codingjoe]

On that note RFC 3966:

>>> import urllib.parse
>>> urllib.parse.urlparse('tel:863-1234;phone-context=+1-914-555:')
ParseResult(scheme='tel', netloc='', path='863-1234', params='phone-context=+1-914-555:', query='', fragment='')

I don't know whether number should be considered a netloc or path either. But at least for SIP it's pretty clear.

[codingjoe]

All I wanted was to create a phone app, but there wasn't a library, so obviously I had to write one... and here we are now 😬