Version "shortened"

[Vangelis forthnet]

On Mon Jul 17 12:32:31 BST 2017, RS wrote: 

> in the LW schedule I referred to 
> Dead Ringers started at 1845 
> but it only ran until 1900.  In other words 
> it was very much shortened from 
> 27min 54s to just under 15 min. 
> The version now available is the full length 
> but it is still being called version shortened. 

Hi Richard :-) 
You are correct; last night when I visited 
http://www.bbc.co.uk/programmes/b08xcwz4/broadcasts 
before posting my answer to Alastair's query, 
the first broadcast entry was:  

Fri 14 Jul 2017 18:45 BBC Radio 4 LW 

that's why I mentioned it in my post; today I see that 

Fri 14 Jul 2017 18:30 BBC Radio 4 FM 

has been (belatedly) added as first broadcast date!
As you said, browsing 
http://www.bbc.co.uk/radio4/programmes/schedules/lw/2017/07/14
one can clearly see the time of broadcast 
(and expected duration) of the LW version; 
clicking the DR entry in that schedule takes you to 
http://www.bbc.co.uk/programmes/b08xcwz4
where its duration is advertised as 15 minutes (on the page). 
But currently the embedded player streams 
the full episode (and GiP does fetch likewise...).

What is still odd is that if one browses 
http://www.bbc.co.uk/radio4/programmes/schedules/fm/2017/07/14
one can see that DR (as you said) was broadcast 
for ca. 30min starting 18:30 BST, but the link there 
again takes you to the same episode page 
http://www.bbc.co.uk/programmes/b08xcwz4
where it is (again) advertised as being only 15min long!

Me thinks this is a c*ckup on the beeb's end! 
They should've created two different pids 
(or even two different vpids), one for the abridged 
(on LW) and another one for the full (on FM) version. 

But I think I (possibly even you) have entered OT territory; 
those comments should've been communicated 
to the BBC; GiP can only grab what's put online 
by the BBC; if they messed things up, it's not a GiP's fault...

> Another curiosity is that estimated mode sizes in --info 
> show much larger sizes for DAF modes 
> than the corresponding HAF and HLSAAC modes.

I think the keyword here is "estimated"; I did 

get_iplayer --type=radio --pid=b08xcwz4 -i | findstr modesizes

and, e.g. 

hafhigh*=36MiB, dafmed*=19MiB, hafstd*=15MiB, hafmed*=11MiB

But both dafmed & hafmed will yield a 19.3 MiB m4a file 
(HE-AACv1 at 96kbps) - so dafmed estimate is the most accurate; 
hafstd will, instead, produce a 25.8 MiB m4a file 
(AAC LC at 128kbps), a big diff from the initial estimate of 
just 15MiB. 

 I don't think something could be done codewise 
to improve the size estimations beforehand (i.e. without actually 
starting a downlod); GiP already tries to make best use 
of whatever size info is advertised in the relevant mediaselector 
pages, when/if that is present; especially for HLS (haf, hlsaac, 
hvf) modes, size has to be calculated based on average stream 
bitrate and sometimes expected and actual bitrate do differ 
(especially in video files). 

 Another reason we're seeing here larger relative diffs 
is because of the small size of the audio file to begin with; 
if it were a big video file of, say, 600+ MiB, then a few 
MiBs offset wouldn't strike that bad...

Best regards, 
Vangelis.