[FFmpeg-soc] [soc]: r5825 - mms/mmsh.c

[zhentan feng]

Hi

On Mon, Jun 7, 2010 at 11:26 AM, Josh Allmann <joshua.allmann at gmail.com>wrote:

> On 6 June 2010 18:38, zhentan feng <spyfeng at gmail.com> wrote:
> > Hi
> >
> > On Mon, Jun 7, 2010 at 1:36 AM, Martin Storsjö <martin at martin.st> wrote:
> >
> >> Hi Zhentan,
> >>
> >> On Sun, 6 Jun 2010, spyfeng wrote:
> >>
> >> > Author: spyfeng
> >> > Date: Sun Jun  6 19:17:38 2010
> >> > New Revision: 5825
> >> >
> >> > Log:
> >> > add primary version for mmsh.c which can't pass the compile.
> >> > TODO:
> >> > 1, parse the asf header.
> >> > 2, finish mmsh_read().
> >> > 3, compile and fix bugs.
> >> > 4, extract common codes with mmst.c
> >>
> >> It seems you're using the tcp protocol directly here - would it be
> >> possible to use the http protocol instead, with a few modifications?
> That
> >> would reduce the code duplication. As you've seen in other mails here,
> >> Josh is adding some customizations to the http protocol, to allow
> setting
> >> custom headers. Is that enough in order for you to use it, or do you
> need
> >> some other modifications, too?
> >>
> >>
> > yes. you are right. I intended to code based on http.c.
> > but I am not sure how to modified it exactly, it may pollution http.c.
> > so I decided to make a wheel from tcp(it's not complicated :p), and make
> > mmsh as a separated file.
> > once I make mmsh.c works, I have already knew exactly what I should
> modify
> > the code.
> > then I will merge it use http.c.
> >
>
> I took a quick look, nice work. It seems one of the things you need,
> that http can't do currently, is the ability to custom-parse response
> headers. RTSP-HTTP actually could use that too, for full spec
> compliance. Other than that, it seems that http.c with my patches
> should do most of what you need.
>
> Just curious -- does MMSH require HTTP/1.0? Are servers particularly
> strict about this? I ask because 1.0 is "required" for RTSP-HTTP, but
> lavf sends out 1.1 headers. Hasn't manifested itself as a problem,
> though.
>
> no. the spec says that HTTP/1.0 and HTTP/1.1 are both should be ok.


>  > I see you set a custom User-Agent field - is that only to minimize the
> >> differences, or is it strictly needed? If it is needed, this is a point
> >> for allowing overriding existing headers with the custom headers in
> Josh's
> >> patchset.
> >>
> >> I am sorry I haven't read Josh's patch, I'll check it and learn
> something
> > from it.
> > thanks a lot.
> >
>
> If you have questions about my patches, I'd be happy to answer them.
>
>
thanks.
have your patch been accepted in trunk?
or which thread in the mail list I should follow with it?
and what about your nickname in IRC?
my nick is "spyfeng" :)

zhentan
-- 
Best wishes~