Odd fetching of radio

[Vangelis forthnet]

On Thu Feb 18 12:04:59 GMT 2016, Jim web wrote:

> OK, I'll note fuller details next time I try it.
> (snip)
> I looked at the content of the get_iplayer file
> but could only see it described as 2.95-dev.
> Afraid I'm not sure where I should look for more
> specific details of the version.
> However I fetched and set it up on 9th Jan
> if that helps at all.

 ... On Windows I am doing ad hoc downloads
of specific revisions (commits) of the main perl
script (get_iplayer) in the develop branch,
as can be found at:

https://github.com/get-iplayer/get_iplayer/commits/develop/get_iplayer

then rename downloaded script to include the
commit hash. E.g., if I downloaded

https://raw.githubusercontent.com/get-iplayer/get_iplayer/232df93d14bb27e4a69b28c01c97603de3de5bf3/get_iplayer

I would rename it to

get_iplayer-295dev-g232df93.pl

place it inside the GiP installation directory,
then launch the Get_iPlayer shortcut and invoke
the dev script as

perl get_iplayer-295dev-g232df93.pl <options>

(I know you're not on Windows, just mentioning
this as a tip for Win users...).
This way, I know exactly which snapshot of the
dev version I am testing...

If you just fetched HEAD from the develop branch
on the 9th of Jan, then it is a safe bet to
assume it was the snapshot with commit hash

d0d2587e8c4dafd3cea07c354e6ab8aa12286740

(https://github.com/get-iplayer/get_iplayer/commit/d0d2587 ,
"Add shoutcast to mode sort") which was pushed
on 20160109013036Z; the next commit was not pushed
until 20160110192128Z. So you should be on

get_iplayer-295dev-gd0d2587

If you know basic git, then the following
most informative blog post should help you
identify the git revision you are currently
running:

http://blog.endpoint.com/2014/11/finding-specific-git-commit-at-point-in.html

On Thu Feb 18 11:33:04 GMT 2016, Jim web wrote:

> I did a comparision this morning
> between using hls for TV fetching
> and DASH for radio.

 With respect, this is a classic case
of apples and oranges. Some facts:

[Do not compare GiP 2.94 and 2.95dev.
The former uses FFmpeg for HLS, the latter
uses by default for both HLS & DASH a native
pure perl downloader (can switch to FFmpeg
for HLS only, as FFmpeg does not do DASH).]

If on 2.95dev, comparisons between HLS TV
and HLS/DASH radio lack a common ground, because:

1. There are various HLS based tvmodes; the
"older" ones which are being served from
Akamai CDN only (hostname "cp401489-vh.akamaihd.net")
and the newer Video Factory ones, which are
being served from either Akamai (hostname
"vod-hls-uk-live.edgesuite.net") or Limelight
CDN (hostname "vod-hls-uk-live.bbcfmt.vo.llnwd.net").

2. HLS radio-on-demand is being served only
from Akamai CDN, but from a different hostname
("cp401492-vh.akamaihd.net") to the one used for TV.

3. Likewise, comparisons between HLS TV and DASH
radio or between HLS radio and DASH radio are
indeed a moot point, since DASH radio is being
served (as you've found out) by either Akamai
CDN (hostname "vod-dash-uk-live.edgesuite.net")
or Limelight CDN (hostname
"vod-dash-uk-live.bbcfmt.vo.llnwd.net").

 Different CDNs, different servers for the
same CDN (with different configurations/
traffic loads / distances from your physical
location amongst them), add to the above
undefined variables pertaining to your own
internet connection, you easily get the
picture that the achieved download speed
at any given time is the result of a
stochastic process...

 You may continue to perform speed measurements,
if you so wish, but IMHO they bear little
statistical value (but then again I wasn't
that good in Statistics/Quantum Mechanics
in my Uni days, long gone they are...).

Kind regards,
Vangelis.