Re: Review request: python bindings for libiscsi

[Martin Sivak <msivak redhat com>]

> It is a lot more code, and notice how the libiscsi_auth_info struct
> contains a 
> method field and an union, currently there is only chap support, but
> in the 
> future there will also be radius and who knows which other methods.
> The way you 
> propose lies madness when the "polymorphism" of the libiscsi_auth_info
> struct 
> really starts to get used, this is why python code simply should not
> override 
> native code, in this case there is a one on one mapping between python
> 
> attributes and structure members, that is not always necessary the
> case, in 
> sometimes parts of the structure may be opaque, in which case your
> solution 
> simply will not work.

But, your code uses native C structure in an object, which doesn't hold it. This will become problem immediately when somebody does anything pythonish. You cannot create module for a language and expect people to not use it's features.

Create better encapsulation, using some form of dictionary if you wish, so the code gets simplier, just do not work with C internals of an object in different object.

> As I understand this is something which has been widely discussed in
> python 
> land in general, and there is no consensus. As such I'm going to keep
> things 
> doing my way, esp as when it comes to code less is more.

This is discussion was only about the isinstance call, the rest is problem everywhere

Martin