[Yum-devel] enable yum to pass http data to repository

[Bryan Mills]

Michael Stenner wrote:

>On Sat, Nov 19, 2005 at 09:34:19AM -0500, Bryan Mills wrote:
>  
>
>I think what you propose is too complex for yum.  I think there are
>too many config options.  All of this stuff would really be used only
>be the extreme expert who is setting up some wacky custom
>authentication system and so it's reasonable to give them a more "raw"
>interface.
>  
>
My only concern is that we are doing url parsing magic that may or may 
not be apparent to the end user.  My general feeling is to avoid all url 
munging and just store the variables that we need which unfortunately 
means there will be more config parameters.

>I've looked over your patches.  For the most part, I have no major
>objection.  I agree that there's a bit of a short-coming in the mirror
>code because you can't append stuff onto the end (like a query).
>However, I'd prefer to make it a bit more general and not try to make
>it too smart.  I don't like the idea of having urlgrabber cleverly
>split up the url and insert stuff in-between.  I also don't really
>like the "query" option which then tries to decipher whether you're
>using GET or POST.  I'd much rather just have an "append" option which
>tacks stuff onto the end of the url.  It can be a query, more
>directory structure, or whatever.  You can still use the "data" arg
>for passing POST data.
>  
>
So on the yum side I'll do the magic url parsing and create the grabber 
objects and passing those into the mirror object directly.  If lots of 
people started using urlgrabber for mirror lists of this type (meaning 
they have http post/get auth requirements) then at that point it should 
be moved it into urlgrabber.

>How does that sound?  You would still be able to do everything you
>want: you use the "append" arg for GET queries and the "data" arg for
>POST queries.
>  
>
Fine with me, do you want me to work up a patch?  Maybe I can get some 
turkey gravy on them.

-bryan