> Date: Mon, 15 Dec 2008 10:48:10 +0100 > From: [email protected] > To: [email protected] > Subject: Re: [libav-user] MPEG4 streaming questions (connection time, > authentication...) > > sebarnolds [Seb / Frip] wrote: > [...] > > - I must enable anonymous viewer login in the IP camera settings because > > authentication is supported. From what I've seen on another post > > (http://lists.mplayerhq.hu/pipermail/libav-user/2008-December/001762.html), > > this is still not supported. Is this right ? > > AFAIK, support for RTSP authentication is still missing. > > > > I looked a bit and it seems it would be easy to support > > rtsp://login:p...@ip/stream urls as the http protocol already support that. > > Again, am I right ? > > If you know the RTSP protocol and the authentication mechanism, > and have some knowledge about the rtsp.c internals, then it should > be easy. > > > > I could do this but I would like to have some input from you as I am not a > > FFMPEG expert. > > Well, I do not know the details of the authentication mechanism. > If you post some details (what needs to be done, how the RTSP setup > phase needs to be modified, etc...) maybe I can provide some hints > about where to implement the needed features in rtsp.c (maybe if you > plan to implement features for the rtsp demuxer the ffmpeg-devel > mailing list can be more appropriate).
Well. If I remember correctly, all we have to do is to reproduce the authentication code from HTTP to RTSP. Currently, when trying to connect to rtsp://login:[email protected]/mpeg4, it doesn't see the login:password@ part (I don't remember if it just ignores it or if it drops it). It needs to parse it, remove it from the URL and include a WWW-Authenticate field with each RTSP request. > > > > - The first frames of the stream that I display are not complete (display > > errors...) and this has a big negative visual impact. If I am not mistaken > > this is because FFMPEG did not receive already a keyframe. Is there a way > > to be sure to have received one to avoid the incomplete display ? > > Not sure about this... In theory, I think you can check the > key_frame field in AVFrame (and avoid displaying any frame until > the first key frame has been received), but I never tried it. > Thank you, it seems to do the trick. Anyway it was a mistake on my side as the first frame read by FFMPEG seems to always be a keyframe. > > > - Calling av_find_stream_info takes a lot of time (up to 4-5s) and this is > > something that I would like to avoid as, as I said, I need to be able to > > switch quickly from a camera stream to another one ? I have found out that > > I could skip the call to av_find_stream_info when using MJPEG streams but > > not with MPEG4 streams. Anyone can tell me one I could reduce that delay (1 > > or 2s would be more acceptable) ? > > Try playing with max_analyze_duration; it should allow to reduce > this startup delay... > Not sure about this one. Reducing it from 3s to 1s didn't seem to help. I put it to 0 and now it seems to work. > > Luca > _______________________________________________ > libav-user mailing list > [email protected] > https://lists.mplayerhq.hu/mailman/listinfo/libav-user Thank you very much for your help. Now, my only remaining problem is the RTSP authentication. I'll have to see if my boss gives me some time to implement it. Sebastien _________________________________________________________________ Send e-mail faster without improving your typing skills. http://windowslive.com/Explore/hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_speed_122008 _______________________________________________ libav-user mailing list [email protected] https://lists.mplayerhq.hu/mailman/listinfo/libav-user
