Re: [Gnash-commit] /srv/bzr/gnash/trunk r11136: migrated rtmp netstream and netconnect classes from rsavoye's local branch and fixed various test cases in misc-ming.all

[Rob Savoye]

On 06/17/09 08:12, Benjamin Wolsey wrote:

> irrelevant. I am talking about the way the NetConnection_as object
> interfaces with whatever network code is used. The NetConnection code
> can be _and should be_ substantially the same as it was, regardless of
> whether libnet / libamf is used or the current code and curl.

  Well, I'm not 100% sure exactly what code you mean here. I did change 
substantially how the existing code worked internally. The lack of any 
existing test cases beyond the minimal ones in actionscript.all makes it 
difficult to test for breakage. All the new test cases pass, and the old 
code couldn't pass any of them, so I assumed it's fine. So I'd love to 
know exactly what code you are referring too.

  I still have a very difficult time understanding why code that works 
only with one web site, namely OpenStreetMap, is considered in better 
shape than code that handles much more. I have even more difficulty 
understanding why code with minimal test cases is also considered 
"better" than code with thousands of tests.

> This's why I said it is trivial to integrate libnet into the existing,
> good design. It should be relatively quick, simple, and not risk losing
> weeks of testing and work.

  The current version was well tested, even it the migration of it was not.

        - rob -