Thanks everyone's help. I will change my device and let it provides profile-level-id and level-asymmetry-allowed.
On Tuesday, March 10, 2015 at 11:47:25 PM UTC+8, Byron Campen wrote: > On 3/10/15 7:09 AM, Randell Jesup wrote: > > On 3/10/2015 9:25 AM, [email protected] wrote: > >> Hi Byron, > >> > >> I add "a=fmtp:97 level-asymmetry-allowed=1" in the answer sdp, and > >> then it works. > >> But actually, I don't think "fmtp" is a necessary attribute in sdp, > >> right? > >> > >> Why FF37 does force other devices sending "fmtp" in answer? > > > > a=fmtp is not generally *required* - however, for H.264, while there's > > a default for packetization-mode (0), you should not assume a default > > for profile-level-id or fail to provide it, even though in theory you > > can (the default is 420010, which only allows a max resolution of > > 176x144@15fps). > > > > It is an exceedingly minor bug that our code apparently requires an > > ftmp line for H.264. No reasonable H.264 device should fail to > > provide it. > > > If level-asymmetry-allowed is not specified, we must assume the > value is 0, and if profile-level-id is not specified, we must assume > 420010 as jesup points out. Since these parameters are specified > independently by each end, if we offer something other than 420010, this > is a mismatch according to the spec. You can argue for bending/breaking > the spec in this case perhaps (where the missing fmtp comes in the > answer, and the offerer already specified that it supports > level-asymmetry, on the assumption that the answerer must have been ok > with level asymmetry also because otherwise it would have not used the > codec). > > The reason for the behavior change is that we have completely > rewritten our JSEP engine, and one of the things the old one did was > break spec and assume a missing level-asymmetry-allowed param had a > default value of 1, instead of 0. > > Best regards, > Byron Campen _______________________________________________ dev-media mailing list [email protected] https://lists.mozilla.org/listinfo/dev-media
