[186039] in North American Network Operators' Group
Re: Binge On! - And So This is Net Neutrality?
daemon@ATHENA.MIT.EDU (Owen DeLong)
Mon Nov 23 18:25:56 2015
X-Original-To: nanog@nanog.org
From: Owen DeLong <owen@delong.com>
In-Reply-To: <20151123225817.314A03D77872@rock.dv.isc.org>
Date: Mon, 23 Nov 2015 15:22:45 -0800
To: Mark Andrews <marka@isc.org>
Cc: NANOG <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
> On Nov 23, 2015, at 14:58 , Mark Andrews <marka@isc.org> wrote:
>=20
>=20
> In message <E24772E7-A95B-4866-9630-2B1023EBD4FD@delong.com =
<mailto:E24772E7-A95B-4866-9630-2B1023EBD4FD@delong.com>>, Owen DeLong =
write
> s:
>>=20
>>> On Nov 23, 2015, at 14:16 , Christopher Morrow
>> <morrowc.lists@gmail.com> wrote:
>>>=20
>>> On Mon, Nov 23, 2015 at 5:12 PM, Owen DeLong <owen@delong.com> =
wrote:
>>>> Except there=E2=80=99s no revenue share here. According to =
T-Mobile, the
>> streaming partners
>>>> aren=E2=80=99t paying anything to T-Mo and T-Mo isn=E2=80=99t =
paying them. It=E2=80=99s kind
>> of like zero-rating
>>>> in that the customers don=E2=80=99t pay bandwidth charges, but =
it=E2=80=99s different
>> in that the service
>>>> provider isn=E2=80=99t being asked to subsidize the network =
provider (usual
>> implementation of
>>>> zero-rating).
>>>=20
>>> equal exchange of value doesn't have to be dollars/pesos/euros
>>> changing hands right?
>>> -chris
>>=20
>> Sure, but I really don=E2=80=99t think there=E2=80=99s an exchange =
per se in this case,
>> given that T-Mo
>> is (at least apparently) willing to accommodate any streaming =
provider
>> that wants to
>> participate so long as they are willing to conform to a fairly basic =
set
>> of technical criteria.
>=20
> No. This is T-Mo saying they are neutral but not actually being so.
> This is like writing a job add for one particular person.
>=20
> Its just as easy to identify a UDP stream as it is a TCP stream.
> You can ratelimit a UDP stream as easily as a TCP stream. You can
> have congestion control over UDP as well as over TCP. Just because
> the base transport doesn't give you some of these and you have to
> implement them higher up the stack is no reason to throw out a
> transport.
Are there a significant number (ANY?) streaming video providers using =
UDP
to deliver their streams?
I admit I=E2=80=99m mostly ignorant here, but at least the ones I=E2=80=99=
m familiar with all use TCP.
Further, it depends on how you define a stream=E2=80=A6
If a stream is a conversation between two particular endpoints using =
consistent
port numbers, then sure, it=E2=80=99s (somewhat) easy to identify, =
except=E2=80=A6
OTOH, if a stream is considered all of the packets involved in a =
particular user
watching a particular video, then depending on implementation, this =
could be much
harder to identify over UDP than TCP.
For example, if the stream is delivered via a torrent-like delivery =
system over UDP,
it could be very hard to identify that all the various seemingly random =
UDP packets
are part of that particular video delivery.
If the requirements were specific enough that they matched a =
particularly small
subset of video delivery services, then I might agree with you. In this =
case, they
seem to have been written more from the technical limitations of =
T-Mobiles current
ability to identify the traffic than targeted at a specific service.
For example, I seriously doubt that video delivered from =
http://us-st.xhcdn.com/swf/ <http://us-st.xhcdn.com/swf/>=E2=80=A6
is likely to be among their =E2=80=9Ctarget candidates=E2=80=9D. =
Nonetheless, it does appear that if
xhcdn chooses to apply under the program, they wouldn=E2=80=99t have any =
trouble meeting
the requirements. (if you want to review the kind of videos hosted on =
xhcdn, visit
their client www.xhamster.com <http://www.xhamster.com/>. Warning=E2=80=A6=
NSFW)
You can make all the claims you want about how they should have or could =
have
implemented this, but unless you have evidence that the issue is =
actually an
attempt to circumvent the intent of net neutrality and not merely a =
technical
limitation of their particular implementation, then I really don=E2=80=99t=
think you have
a basis for your claim above.
Owen
